This action might not be possible to undo. Are you sure you want to continue?
DOI 10.1007/s1099201191826
A Tractarian Universe
Albert Visser
Received: 30 July 2010 / Accepted: 17 February 2011
© The Author(s) 2011. This article is published with open access at Springerlink.com
Abstract In this paper we develop a reconstruction of the Tractatus ontology.
The basic idea is that objects are unsaturated and that Sachlagen are like
molecules. Bisimulation is used for the proper individuation of the Sachlagen.
We show that the ordering of the Sachlagen is a complete distributive, lattice.
It is atomistic, i.e., each Sachlage is the supremum of the Sachverhalte below
it. We exhibit three normal forms for Sachlagen: the bisimulation collapse, the
canonical unraveling and the canonical bisimulation collapse. The first of these
forms is unique modulo isomorphism, the second and third are simply unique.
The subset ordering on normal forms of the second and third kind reflects the
ordering of the Sachlagen.
Keywords Wittgenstein · Tractatus · Ontology · Bisimulation · Unraveling
1 Introduction
Unsaturatedness is the hallmark of Tractarian objects. This, at least, is the inter
pretation of Tractatus that we will pursue in this paper. We develop a model
of the Tractatus ontology, in which occurrences of unsaturated objects click
Dedicated to Roel de Vrijer on the occasion of his 60th birthday.
A. Visser (B)
Department of Philosophy, Utrecht University,
Janskerkhof 13a, 3512 BL Utrecht, The Netherlands
email: albert.visser@phil.uu.nl
A. Visser
together to form saturated Sachlagen and Sachverhalte.
1
The Sachverhalte are
atomic Sachlagen—in a sense of atomic that will be explicated.
An important point of the paper is the analysis of sameness of Sachlagen.
This is an issue that Wittgenstein did not address, probably because he never
attempted to build a concrete model of his Logical Space. As will be explained
in the paper we opt for bisimularity as the appropriate analysis of sameness.
This means, very roughly, that two Sachlagen are the same as they cannot be
distinguished from an internal structural point of view.
1.1 Saturated Versus Unsaturated
Frege insisted that only in the context of a sentence does a word have meaning
(the context principle). Now consider the sentence Roel is happy. We
can view the predicate expression is happy as obtained by removing the
name Roel from the sentence. So the predicate expression really is: ( ) is
happy. We can now see that the predicate expression is ergänzungsbedürftig
or unsaturated: it needs supplementation to form an expression that is mean
ingful in its own right, to wit the sentence. This Ergänzungsbedürftigkeit is, in
Frege’s set up, inherited on the semantic level by the predicate itself.
But what about referring expressions? What about Roel? It seems to me
that if we were really serious about the context principle, we should say that
Roel was really Roel ( ), and consequently that the object corresponding
to the name is unsaturated. If we forget philosophy of language for a moment,
we could say that, in this view, properties and relations can only have reality in
the objects having them, but, conversely, that objects also only have reality in
the properties they have and in the relations they stand in. Such a view would
blur the distinction between object and property/relation.
The tradition has refrained from taking the step towards unsaturated ob
jects. E.g., in Montague Grammar, we have a basic type of objects and a
basic type of propositions/truth values. All other types are derived from these.
The objects of the basic types can be viewed as the saturated entities in this
framework: these are the entities that are not functions.
Wittgenstein, in my reconstruction, insisted that objects are ergänzungs
bedürftig too. This is reflected in the fact that they do not occur independently.
Here is a telling quote:
2
2.0131 Der räumliche Gegenstand muß im unendlichen Raume liegen.
(Der Raumpunkt ist eine Argumentstelle.) Der Fleck im Gesichtsfeld
muß zwar nicht rot sein, aber eine Farbe muß er haben: er hat sozusagen
den Farbenraum um sich. Der Ton muß eine Höhe haben, der Gegen
stand des Tastsinnes eine Härte, usw.
1
I will leave “Sachlage” and “Sachverhalt” untranslated. They are Tractarian expressions with
very special meanings. Still, the Pearce & McGuiness translation of “Sachlage” as situation, and of
“Sachverhalt” as state of af fairs, does reflect my interpretation of these words reasonably well.
2
We use the Pearce & McGuiness translation [9] throughout the paper.
A Tractarian Universe
2.0131 A spatial object must be situated in infinite space. (A spatial point
is an argumentplace.) A speck in the visual field, though it need not be
red, must have some colour: it is, so to speak surrounded by colourspace.
Notes must have some pitch, objects of the sense of touch some degree of
hardness, and so on.
In short, happiness cannot occur without someone being happy, but Roel
cannot occur without being in some mood.
This reading finds some support in the exegetic literature. For example, here
is the discussion by Max Black, in his classical [2], p 13. He says:
Frege took a decisive step forward when he introduces a radical distinc
tion between ‘functions’ and ‘objects’. The former being ‘incomplete’, ‘in
need of supplementation’, ‘unsaturated’, have to be symbolized by words
of a peculiar and distinctive sort (‘function names’), having ‘gaps’ that
need top be filled by names of objects. Wittgenstein went still further:
for him, it might be said, all simple symbols (names) were unsaturated in
something like Frege’s sense. Names occur only in association with other
names, and have no reference except in the propositional context (3.3); in
elementary propositions they grip one another, without intermediaries,
like links of a chain (cf. 3.14); and their esse is to be eligible for such
concatenation (3.203+3.21+2.0123).
We note that Black is discussing the unsaturatedness of names, not of
objects. We take the unsaturatedness of names to be reflected in the unsat
uratedness of objects, or, if you wish, the unsaturatedness of names to reflect
the unsaturedness of objects. (Black refers to 2.0123 which is about objects.)
The above discussion takes our ordinary understanding of world and lan
guage as starting point and reasons from there about what the tractarian
objects could be. If we start from the standpoint of Wittgenstein’s ontology, we
just have unsaturated objects.
3
There are no relations and traditional objects.
Since Tractarian objects are not given to us in ‘surface language’, we could
easily imagine that for some of these objects it is impossible to say whether
they are more like traditional objects or traditional relations. They simply need
not have traditional counterparts.
Remark 1.1 Nominalism versus Realism: There has been some discussion
whether universals like R are objects in Wittgenstein’s sense. Do they occur as
objects in configurations, or do they subsist in the way objects click together?
The positive answer to this question can be labeled realism, the negative
answer nominalism.
Tractatus 3.1432 seems to speak against the objectual understanding of
universals. On the other hand, our leading idea of a rapprochement between
objectsquaparticulars and universals because they are all just unsaturated,
strongly suggests an equal treatment of a and R.
3
I thank Jesse Mulder for stressing this point to me.
A. Visser
We have already stressed that the question concerning traditional objects
and relations, does not arise when we start from the ontology. Since our
framework just models the ontology, it is neutral between realism versus nom
inalism. However, the motivation of our framework, looking from language to
ontology, fits better with the realist option.
It is more clumsy to interpret familiar patterns of predication nominal
istically. Consider, e.g., Sachverhalte [P
0
a], [P
1
a], . . . , of the traditional
subjectpredicate form, where the P
i
are supposed to be different. If in the
Sachverhalte nothing corresponds with the P
i
, the differences between the
Sachverhalte must be accounted for as different ways in which a is linked to
itself. For example, we could suppose that a has two poles p, q, where both
poles may click together both with themselves and with each other.
4
This
would give us two possible Sachverhalte only involving one occurrence of a:
one where each pole clicks with itself en one where the poles click with each
other. More generally an object that has n poles that can click together both
with themselves and with each other, can form a oneobject Sachverhalt in n!
ways. So there are precisely n! unary predicates applicable to a. Of course, we
can make the different poles of a of different types, so that more interesting
patterns of clicking together become possible. Thus, we can raise the number
of possible properties of a by working with more poles that fit together. All
this seems rather unattractive. It feels like a hack.
See for a discussion of exegetical matters concerning the nominalism versus
realism issue, e.g., the chapter on Problèmes ontologiques in [7], pp. 68–84.
1.2 Objects in Multiplicity
How precisely do the objects form a chain or configuration? We will assume
that configurations are not setlike. If they were setlike, how could we explain
the difference between [aRb] and [b Ra]? They are more like molecules where
the objects are connected via links or poles.
Consider a Sachverhalt of the form[aRa]. Wittgenstein’s discussion of iden
tity could carry some suggestion that he would not allow configurations of this
form. However, there is a powerful reason not to disallow such configurations,
to wit intersubstitutivity of objects of the same logical form in Sachverhalte.
The notion of logical form does not necessarily imply a substitution principle,
but it certainly does cohere well with a substitution principle. We would like
to demand the following: if a and b have the same logical form, then we can
always substitute a for b in a configuration. Thus [aRa] can be ‘derived’ from
[aRb]. How can we think about [aRa]? One idea is that a has two poles, say
p and q. The pole p combines with the first pole of R, the pole q combines
with the second pole of R. Saturatedness of [aRa] dictates that a does not have
more poles. But then consider [Pa]. Here a can have only one pole. Should we
4
We say that the objects are connected via poles. This will be discussed in more detail in
Section 1.3.
A Tractarian Universe
allow a to have a varying number of poles? But if this number varies, it should
vary when we move from[aRa] to [Pa], so what varies here to make a ‘change’
its number of poles?
We consider a variant of our example: presumably both [aRb] and [Pa]
could be Sachverhalte. Since they are supposed to be saturated, in each case,
a has exactly one pole. However, how can this pole be employed twice? This
phenomenon can be explained by the insight that what we find in Sachverhalte
are occurrences of objects. We need not stipulate a varying number of poles to
account for the possibility of both [aRa] and [Pa]: we can say that a has two
occurrences in [aRa]. We note there is some textual evidence that Wittgenstein
thought about objects as occurring in states of affairs. For example, in 2.012,
he says:
2.012 In der Logik ist nichts zufällig: Wenn das Ding im Sachverhalt
vorkommen kann so muß die Möglichkeit des Sachverhaltes im Ding
bereits präjudiziert sein.
2.012 In logic nothing is accidental: if a thing can occur in a state of affairs,
the possibility of the state of affairs must be written into the thing itself.
Thus, we will model our logical spaces by implementing a notion of occur
rence of an object. As we will see, in our framework, occurrences have very thin
identities. Not only are they like indiscernibles, but, in many examples, even
their cardinality is indeterminate.
1.3 How Objects Click Together
The poles are what connects objects to each other. Each pole connects to
another pole. Not all poles may click together. Each pole has a type. Ony
poles with types that match may connect. Which types match is given with the
specification of logical space.
We will assume that the poles of an object are fixed. This is assumed just for
the sake of simplicity. We see no knockdown philosophical argument against
the possibility of a varying number of poles for some objects or relations.
This would give us ‘polyadic objects’. Also, at first sight, it seems pretty
straightforward to generalize our framework to accommodate variable sets of
poles.
5
We will also assume that connection is always a matter of two poles.
6
Again
this is just an assumption for the sake of simplicity. The framework can easily
be adapted to multilinking.
Finally, we will assume that our poles have individuality. They are not
indiscernibles. The reason for this choice is the intuition that the neatest way
5
Varying assemblies of poles are suggested as one possible option for obtaining a more systematic
view of objects as special cases of (possibly) unsaturated configurations. See Section 5, for further
remarks on this matter.
6
This assumption makes our Sachlagen into linear graphs. See Section 1.6, for a reference.
A. Visser
to analyse a firstorder style predication like [Pab], by giving P two poles of
type, say, a and a and b each one pole of type, say, b. Here a and b match.
In Example 2.7, this idea is worked out. So the argument places are given
by the individuality of the poles; they are not coded in the types. Again it is
easy to adapt the framework to have indiscernible poles. E.g., if all poles were
indiscernible, the poles associated to an object would be a multiset of types.
1.4 Sachlagen and Sachverhalte
Sachlagen are like molecules. They are saturated configurations of objects
which are clicked together via their poles in the ways that are allowed by the
nature of logical space. Sachlagen are the primary candidates for existence.
Sachverhalte are special Sachlagen. In the framework they are the smallest
nonempty Sachlagen. A Sachlage is not a set of Sachverhalte, but as we will
see it is fully determined by a set of Sachverhalte.
There is an important interpretive choice in our treatment of Sachlagen. A
Sachlage is a (possible) situation that is depicted in a picture. It is neutral with
respect to existence. An existing Sachlage is a Tatsache or fact. According to
our interpretation, there are no disjunctive pictures. So a sentence of the form
p ∨ q, where p and q are atomic does not depict. Its interpretation is handled
by the truthfunctional semantics, which sits on top of the picture theory.
Sachlagen appear in two roles. We can take them as presenting something of
the form at least this is the case (inclusive role) or as presenting this is the case
and it is all that is the case (exclusive role). In their inclusive role, Sachlagen
are parts of possible worlds and, in their exclusive role, they are full possible
worlds. The actual world is the existing Sachlage in its exclusive role.
A Sachlage in the in the inclusive role corresponds to a conjunction of
atomic sentences, and a Sachlage in the exclusive role corresponds to a con
junction of atomic sentences and negated atomic sentences (together: atomic
literals) in which all possible atomic sentences occur precisely once.
Remark 1.2 We remind the reader of what Wittgenstein says in 2.11:
2.11 Das Bild stellt die Sachlage im Logischen Raume, das Bestehen und
Nichtbestehen von Sachverhalten, vor.
2.11 A picture presents a situation in logical space, the existence and non
existence of states of affairs.
This is somewhat mysterious, because Wittgenstein includes the non
existence of Sachverhalten as something that is depicted. Maybe the existing
Sachverhalte are in black and blue and red and the nonexisting ones in yellow
and pink and sepia? Such a subtle convention, surely would be beyond the
range of simple depiction?
In an alternative reading, Wittgenstein takes the exclusive reading as the
default. Thus, what is depicted would always be the Sachlage in its exclusive
role. In other words, pictures are always full pictures of possible worlds. This
A Tractarian Universe
reading would solve how nonexistence can be pictured, however, it diverges
substantially from the ordinary understanding of picture. Moreover, it seems
improbable that even Wittgenstein would have such a subtle reading of what a
picture does, without commenting upon it.
It would fit my interpretation better if Wittgenstein had omitted the phrase
about nonexistence.
The objects that we produce in our modeling to represent Sachlagen are
neutral with respect to inclusive or exclusive role. A Sachlage will simply be a
saturated configuration of objects.
Between Sachlagen (in the inclusive role) we have the subSachlage partial
ordering. A Sachverhalt is an atom in this ordering. This means that a Sachver
halt is a minimal nonempty Sachlage.
We will prove the theorem that Wittgenstein used but failed to state: the
subSachlage ordering is atomistic, i.e., every Sachlage is the supremum of the
Sachverhalte below it.
Remark 1.3 The framework allows that there are saturated objects. This shows
that it is conceivable that at least some objects are, or, at least, coincide
with Sachlagen. Also both Sachlagen and objects belong to an encompassing
category, say, the conf igurations.
7
For more on configurations, see Section 5.
1.5 When Two Sachlagen are the Same
In the present paper, we are constructing models of logical spaces against the
background of a classical metatheory or modeling medium. As a modeling
medium we employ informal set theory, say, with urelements for items like
objects and poles. We are assuming that the urelements form a set, that the
poles of an object form a set, etcetera. The objects we provide to repre
sent basic items like Sachverhalte and Sachlagen are supposed to represent
Sachverhalte and Sachlagen as they are philosophically intended, but there is
no claim that the items constructed are Sachverhalte and Sachlagen. What is
more the objects we construct contain artifacts of the modeling, aspects that
are not present in the Sachverhalte and Sachlagen as philosophically intended.
One important matter is the question of sameness of two (models of)
Sachverhalte. Consider, for example, an object a with two poles p and q of
type, respectively, a and b. Suppose a and b match. In this case, we could have
one occurrence of a that links with itself. We could also have two occurrences
of a, such that each links with the other. Or we could have three occurrences
of a that link in a circle. We could have an infinity of occurrences of a that
link with the order type of the integers. Are all these possibilities the same or
different?
7
Of course, we could hold that an object is conceptually different from the configuration contain
ing just that object. In fact, I think that is the best way of looking at it.
A. Visser
My intuition about this question can best be explained by introducing the
idea of the inner perspective on a Sachlage. Suppose we had a fairy living in a
Sachlage. What would that be like? The fairy would be at an occurrence of an
object and would be able to travel across the poles to other occurrences. She
would be able to identify those things that have true individuality: the object
where she currently is, the link across which she travels. She would have no
access to the identity of the occurrence: the occurrence is ‘thin’, it is just a
‘multiplicity’. We may compare this idea with the intrinsic notion of curvature
in a geometry that can be understood independently of an embedding space.
My idea is that two Sachlagen are the same if they are the same from the
inner perspective, or even stronger: there is nothing to the Sachlage than
the inner perspective. The outer perspective only exists as an artifact of the
modeling.
Consider two Sachlagen σ and τ. Consider any objectoccurrence (d, i) in
σ and any objectoccurrence (e, j) in τ. Suppose we drop fairy Femke in (d, i)
and fairy Inge in (e, j). We suppose that Femke and Inge can communicate
by mobile phone. We say that (d, i) and (e, j) are internally indistinguishable
or observationally equivalent, if there is no way that Femke and Inge can
find a difference between them by moving from object occurrence to object
occurrence, carefully noting along which poles they travel. So, Inge and Femke
try out paths through their respective Sachlagen and check whether they can
find differences. If they cannot possibly find differences in this process, the ini
tial occurrences are internally indistinguishable. Of course, indistinguishability
implies that d = e.
We note that Femke and Inge are not allowed to leave breadcrumbs.
Leaving breadcrumbs would change the Sachlage, which is impossible—even
for a fairy.
The Sachlagen σ and τ are the same if, for every objectoccurrence (d, i)
in σ, there is an internally indistinguishable objectoccurrence (d, j ) in τ, and
vice versa.
Our intended notion of sameness can be technically explicated using the
well established notion of bisimilarity. This notion is employed in Computer
Science. For example, in Process Algebra it is used to explicate sameness of
processes. It is also employed in Modal Logic to provide a notion of sameness
of Kripke models. It and related concepts provide a good tool for proving
various theorems. The application of bisimulation that is, perhaps, closest to
the use we are making of it in this paper, is the analysis of nonwellfounded
sets. We refer the reader to Peter Aczel’s wonderful book [1].
8
Remark 1.4 There is some abus de langage, in that we call representations of
Sachlagen “Sachlagen”, and representations of Sachverhalte “Sachverhalte”.
It is my feeling that the alternative of using different names, say “pre
8
For a discussion of individuation of sets, see also Adam Rieger’s eminently readable paper [8].
Unfortunately, the Rieger paper does not have so much emphasis on notions of bisimulation.
A Tractarian Universe
Sachverhalte” for representations of Sachverhalte, is so unpleasant that the
awkwardness of our usage is well worth the cost. So, we will use “Sachlage” and
“Sachverhalt” for the items in our modeling, without dividing out bisimilarity.
We will say things like when two Sachlagen are the same, meaning: when two
representations of Sachlagen are bisimilar.
1.6 Linear Graphs
Vincent van Oostrom pointed out to me that my modeling of Sachlagen is
very much similar to linear graphs as studied by Alan Bawden in his paper
Implementing Distributed Systems using Linear Naming (see: http://dspace.
mit.edu/handle/1721.1/7085).
Linear graphs turn out to be almost the same as (models of) Sachlagen.
There are some minor differences. We may translate our concepts into Baw
den’s as follows.
This paper Bawden’s paper Modern graph rewriting
Object Type Type
Objectoccurrence Vertex Vertex
Poleoccurrence Terminal Port
Pole Label –
Type – –
Under this translation, the poles in Bawden’s framework are chosen from a
fixed set. They may be viewed as all having the same type that clicks with itself.
Bawden also uses bisimulation as the proper relation of equality. His
explanation of why this is a good choice is very much similar to mine. Since
Bawden’s study is from 1993, he has clear priority for a number of ideas.
Finally we note that, where Tractarian logical spaces are static, graph
rewriting adds a dynamics and studies processes that modify linear graphs.
1.7 Significance of this Paper
I submit that the study of these logical spaces à la Wittgenstein, has some
interest beyond the historical reconstructive question whether one can make
sense of the ontology of the Tractatus. Even if precisely these spaces are rather
clearly not a good model for natural language semantics, the idea of non
saturated objects is a good one—and one that has not been studied sufficiently
seriously. In this light, what we are doing here can be viewed as a study of
unsaturated objects under radical simplifying assumptions.
In the previous subsection, we discussed the similarity between linear
graphs and Sachlagen. This convergence also shows that we are looking at a
natural idea.
A. Visser
2 Let There be Sachlagen
In this section we develop our (class of) model(s) of the Tractatus universe.
It is important to stress that we are not aiming at a foundational presentation,
building up the Tractarian world and its metatheory in tandem. We simply
employ the usual set theoretic metatheory, say ZF with urelements, and we
assume that the ingredients of Wittgensteinian logical spaces can be taken to
be just sets. It will then, e.g., be a theorem (Theorem 4.3) that the totality of
Sachlagen can be represented by a set.
A logical space is given by:
•
A nonempty set of objects D,
•
A set of types T,
•
A binary matching relation M between types. We demand that M is
symmetric.
•
A function P that assigns to each object a (possibly empty) set of poles.
•
A typing function F on D such that F
d
: P
d
→ T.
Every object comes equipped with a set of poles. Even if the precise
identity of these poles is immaterial, they still are like fixed argumentplaces or
argumentroles. Of course, a more flexible implementation would be possible,
e.g. adding a group of permutations to the set of poles to make certain places
‘the same’, but we will refrain from adding such extra features in the present
implementation.
9
Each pole has a ‘type’. The matching relation tells us which
types may combine.
Two objects d and d
have the same logical form if there is a typepreserving
bijection between P
d
and P
d
.
We turn to the definition of a Sachlage in a given logical space. We model
occurrences by assigning to each object a set of indices.
10
Thus, a Sachlage σ is
given by:
•
A function I from objects to (possibly empty) sets of indices.
•
A binary relation R on O
pole
σ
, where O
pole
σ
, the set of pole occurrences, is
the set of (d, i, p), where i is in I
d
and p is in P
d
. We demand that R is total
(in the sense: ∀x ∃y Rxy), functional and symmetrical and that (d, i, p) R
(d
, i
, p
) implies F
d
( p) M F
d
( p
).
9
Such more refined treatments of argumentplaces are studied in the philosophy of neutral
relations. See, e.g., [4, 5] and [6].
10
This is a good point to note some disanalogies with the molecules of physics. Firstly, we can have
different physical molecules of the same type. In contrast every Sachlage is unique. Secondly, in
physics, the same atom (numerically) cannot occur in different molecules and it cannot occur a
number of times in the same molecule (at the same time). Different atoms of the same type may
occur in different molecules. A type of atom may occur a number of times in the same molecule.
In contrast, a tractarian object may occur in several Sachlagen and may occur a number of times
in the same Sachlage.
A Tractarian Universe
The function I gives us, for each object, the set of occurrences of that object.
The set of occurrences of objects in a Sachlage σ is O
obj
σ
, the set of pairs (d, i),
where i ∈ I
d
. The relation R tells us which poles click together in the Sachlage.
We only allow poles to click together if their types match. The demand that R
is total embodies the idea that Sachlagen are saturated.
One pleasant alternative way to think about R is as a set of arrows α
between objectoccurrences of the form ((d, i), p, q, (e, j )). Here (d, i) is the
domain of α and (e, j ) is the codomain of α. We represent such an arrow as
(d, i)
p,q
−→ (e, j ). We write (d, i)
p,q
—– (e, j ), if we want to stress the symmetry
of R.
It is useful to also have a click relation at hand that ignores the poles via
which occurences of objects are connected. Thus, we define:
•
(d, i)
¯
R (d
, i
) :⇔ ∃p, p
(d, i, p) R (d
, i
, p
).
Remark 2.1 We note that we allow a poleoccurrence to click with itself. This
seems somewhat unnatural. It is like a hand that shakes itself. However,
excluding it will complicate our notion of bisimulation collapse. Moreover,
modulo bisimulation we can always replace such selfclicks by nonselfclicks.
This happens for example in the canonical unraveling. See Section 4.
We turn to the definition of sameness of Sachlagen. We will identify as
many occurrences as is reasonable. We identify all occurrences that we can
consistently identify given that we want their environments to be similar. We
say that B is a bisimulation between σ and σ
iff B is a function on objects that
assigns to each d a relation B
d
between I
d
and I
d
, such that B
d
satisfies the zig
and the zagproperty (see Fig. 1):
•
If i B
d
i
and (d, i, p) R (e, j, q), then, for some j
in I
e
, we have j B
e
j
and
(d, i
, p) R
(e, j
, q) (zigproperty / forward property),
•
If i B
d
i
and (d, i
, p) R
(e, j
, q), then, for some j in I
e
, we have j B
e
j
and (d, i, p) R (e, j, q) (zagproperty / backward property).
Note that, in an alternative presentation, we could view a bisimulation as
a relation between the sets of object occurrences O
obj
σ
and O
obj
σ
. So B of our
official definition would be represented by B
◦
with:
•
(d, i) B
◦
(d
, i
) :⇔ d = d
and i B
d
i
.
Fig. 1 The Zig and the Zag
property
A. Visser
Fig. 2 Zig implies Zag
It is easy to regain B from B
◦
. It will be pleasant to move in a flexible way
between B and B
◦
. The domain of B will be the domain of B
◦
, and, similarly,
for the range.
A bisimulation B is total if, for each object d, the relation B
d
is total. It is
surjective if, for each d, B
d
is surjective. The Sachlage σ is a subSachlage of
σ
, or σ σ
, iff there is a total bisimilation B from σ to σ
. We will write
B : σ σ
, if we want to exhibit the witnessing bisimulation B. The Sachlage
σ is equal to σ
, or σ ≡ σ
, iff there is a total and surjective bisimilation B
from σ to σ
. We will write B : σ ≡ σ
, if we want to exhibit the witnessing
bisimulation B.
Remark 2.2 A surprising aspect of our framework is that the zig property
implies the zag property (and vice versa). Here is the argument. (See Fig. 2).
Consider Sachlagen σ and σ
. Suppose we have the zig property from σ to σ
.
Suppose i B
d
i
and (d, i
, p) R
(e, j
, q). Since σ is saturated, there must be
e
∗
, j
∗
, q
∗
, such that (d, i, p) R (e
∗
, j
∗
, q
∗
). By the zig property, we find that, for
some, j
, we have j
∗
B
e
∗ j
and (d, i, p) R (e
∗
, j
, q
∗
). Since, R is functional, it
follows that (e, j
, q
∗
) = (e, j
, q). So, j
∗
witnesses the backward property.
Each Sachlage σ in its exclusive role represents a possible world. We impose
the reasonable constraint that the existing Sachlagen (in their inclusive role)
are closed under sums. We will show below that this implies that there is a
maximal existing Sachlage.
When viewing σ as a possible world, we say that σ
exists in σ, or σ
is a fact
of σ, if σ
σ.
Here are the basic facts about bisimulations.
Theorem 2.3
a. Let id
σ
be given by: id
σ,d
is the identity relation on I
σ,d
. We have: id
σ
is a
total and surjective bisimulation from σ to σ.
b. For any bisimulation B from σ to σ
, let
˘
B be given by the stipulation that
˘
B
d
is the converse relation of B
d
, for each d. We have:
˘
B is a bisimulation
from σ
to σ.
A Tractarian Universe
c. Let B be a bisimulation from σ to σ
and let B
be a bisimulation from σ
to
σ
. We def ine B; B
by: (B; B
)
d
is the composition in the order of reading
of B
d
and B
d
. We have: B; B
is a bisimulation.
d. Let B be a set of bisimulations from σ to σ
. We def ine
B as given by:
(
B)
d
is the union of the B
d
for B in B. We f ind:
B is a bisimulation.
The proof of the theoremis entirely routine. The theoremgives us immediately
the following corollary.
Corollary 2.4 The relation is a partial preordering with ≡ as its induced
equivalence relation.
Proof For example, if B
0
: σ σ
and B
1
: σ
σ, then (B
0
∪
˘
B
1
) : σ ≡ σ
.
Conversely, if B : σ ≡ σ
, then B : σ σ
and
˘
B : σ
σ.
We define the empty Sachlage, ⊥⊥, by: I
⊥⊥,d
:= ∅, for all d and R
⊥⊥
is the empty
relation. Clearly, ⊥⊥ is the minimal element w.r.t. of . Its ≡equivalence class
of ⊥⊥contains precisely one element. We will show that there is also a maximal
Sachlage . See Theorem 3.11.
We define a Sachverhalt as an atom of . This means that σ is a Sachverhalt
iff, for any σ
σ, we have σ
≡ σ or σ = ⊥⊥.
We end this section by presenting some examples.
Example 2.5 Consider the following logical space BASIC. We just have a single
object ens. The object ens has no poles, i.o.w., P
ens
:= ∅. The space has no
types, and, consequently, the relation M is empty. Our space has two Sachlagen
(modulo ≡). First we have the empty Sachlage nil or ⊥⊥ with I
nil,ens
= ∅ and
R
nil
is the empty relation. Then there is the Sachlage esse or in which ens is
present. Here I
esse,ens
= {0} and R
esse
is the empty relation. Here esse is the
only Sachverhalt of our space.
Example 2.6 Consider the following logical spaces S
0
and S
1
. In both spaces
we just have a single object d. In both spaces, we have P
d
:= { p}. Both spaces
have one type a. We take M
S
0
:= ∅ and M
S
1
:= {a, a}.
Clearly S
0
has only the empty Sachlage ⊥⊥. The poor object d is eternally
blocked from participating is a Sachlage.
The space S
1
has as Sachlagen ⊥⊥ and a second Sachlage solo. Modulo
isomorphism, solo has just two connected representations. The first repre
sentation is the following. The object d has one occurrence, say 0. The click
relation is defined by: R
solo
= {(d, 0, p), (d, 0, p)}. The second representation
looks as follows. The object d has two occurrences, say 0 and 1. We have
R
solo
= {(d, 0, p), (d, 1, p), (d, 1, p), (d, 0, p)}. There are classmany other
representations consisting of sums of disjoint copies of the ones just given.
In Section 4, we will introduce the bisimulation collapse and the canonical
unraveling. Modulo isomorphism, our first representation is the bisimulation
collapse and our second representation is the canonical unraveling.
A. Visser
Example 2.7 Let some first order relational signature be given and some
nonempty set of objects X. We assume that the relation symbols in and
the objects in X are disjoint. We take D the union of and X. For a relation
symbol r of arity n, we take P
r
:= n := {0, . . . , n −1}. For an object x ∈ X, we
take P
x
:= 1. Let F
r,i
:= a and F
x,0
:= b. Let M := {(a, b), (b, a)}.
In this example the Sachverhalte correspond precisely with what we would
take intuitively as Sachverhalte over the signature. E.g., the Sachverhalt rxy,
where r is binary, is modeled as σ, where, for z ∈ X, I
σ,z
= 0 if z = x and z = y,
and I
σ,z
:= 1, if z = x or z = y. Moreover I
σ,s
= 0, if s is a relation symbol not
equal to r, and I
σ,r
:= 1.
Example 2.8 In 4.2211, Wittgenstein writes:
4.2211 Auch wenn die Welt unendlich complex ist, so daß jede Tatsache
aus unendlich viele Sachverhalten besteht und jeder Sachverhalt aus
unendlich vielen Gegenständen zusammengesetzt ist, auch dann müßte
es Gegenstände und Sachverhalte geben.
4.2211 Even if the world is infinitely complex, so that every fact consists
of infinitely many states of affairs and every state of affairs is composed
of infinitely many objects, there would still have to be objects and states
of affairs.
In this example, we showthat one may imagine Sachverhalte that consist not
of infinitely many objects, but of infinitely many occurrences of two objects.
In our example we have an f inite logical space with an inf inite Sachverhalt.
Consider a logical space with two objects d and e. The first object d has one
pole, to wit p. The second object e has two poles, to wit q and r. The types are
a and b. We set F
d
p := a, F
e
q := b, F
e
r := a. We take M:= {(a, b), (b, a)}.
We now consider the Sachlage that is given as follows.
•
I
d
:= {∗}, I
e
:= ω,
•
The relation R is generated by the following stipulations:
(d, ∗, p) R (e, 0, q), (e, 0, r) R (e, 1, q), . . . , (e, n, r) R (e, n +1, q), . . .
Thus, our our Sachlage looks like this:
(d, ∗)
p,q
(e, 0)
r,q
(e, 1)
r,q
. . . (e, n)
r,q
(e, n+1) . . .
It is easily seen that this example indeed gives us a Sachlage (even a Sachver
halt) and that the maximal bisimulation of this Sachlage and itself is the
identity, so the infinity of occurrences is ‘real’.
We leave it as an exercise to the reader to construct an infinite Sachlage in
a logical space with only one object.
A Tractarian Universe
3 The Ordering of the Sachlagen
In this section we study the ordering . We show a.o. that it is, modulo ≡, an
atomistic, complete distributive lattice. The results of this section often have
alternative proofs using the normal form theorems of Section 4.
3.1 Basic Facts
To develop some feeling for what a Sachlage looks like, we will first prove
some basic facts. To formulate these we give some basic definitions.
•
A set of objectoccurrences X is closed in σ, if, whenever (d, i) ∈ X and
(d, i)
¯
R
σ
(e, j ), then (e, j ) ∈ X. If the Sachlage is clear from the context,
we will simply say that X is closed.
•
Suppose X is closed in σ. We define σ X or σ
X
as follows.
– I
σX,d
:= {i ∈ I
σ,d
 (d, i) ∈ X},
– (d, i, p) R
σX
(d
, i
, p
) :⇔ (d, i, p) R
σ
(d
, i
, p
),
where i ∈ I
σX,d
and i
∈ I
σX,d
.
Clearly, σ
X
is again a Sachlage.
•
σ τ iff, for all d, I
σ,d
⊆ I
τ,d
and R
σ
⊆ R
τ
.
•
E
σ
is the transitive, reflexive closure of
¯
R
σ
. Clearly, E
σ
is an equivalence
relation. We call the E
σ
equivalence class of (d, i) in O
obj
σ
: [d, i]
σ
. A
Sachlage σ is connected if it has at most one E
σ
equivalence class. We write
σ[d, i] for σ [d, i]
σ
. It is easily seen that σ[d, i] is a connected Sachlage.
•
Suppose B is a bisimulation between σ and τ (not necessarily total or
surjective). Suppose σ
σ and τ
τ. We define B
σ
,τ
by i B
σ
,τ
,d
j
iff (d, i) ∈ O
obj
σ
, (d, j ) ∈ O
obj
τ
and i B
d
j. It is easy to see that B
σ
,τ
is a
bisimulation between σ
and τ
.
•
Suppose B is a bisimulation between σ
and τ
(not necessarily total or
surjective). Suppose σ
σ and τ
τ. We write B
σ,τ
for the bisimulation
between σ and τ such that, for every d, B
σ,τ
d
has the same graph as B
d
. In
other words B
σ,τ
is B, but for the fact that we consider it as a bisimulation
between other Sachlagen.
We have τ σ iff τ = σ O
obj
τ
. Moreover, τ σ implies that τ σ, since
the identical embedding of the sets of indices for an object forms a total
bisimulation.
Clearly, σ is divided in (possibly zero) nonempty connected parts, the E
σ

equivalence classes. Some of these E
σ
equivalence classes may be bisimilar
to each other. We will see that each E
σ
equivalence class Y corresponds to
a Sachverhalt. The classes bisimilar to Y are bisimilar copies of the same
Sachverhalt. Here is a first lemma.
Lemma 3.1 Suppose B is a bisimulation between σ and τ.
i. The domain of B
◦
is closed in σ and the range of B
◦
is closed in τ.
A. Visser
ii. Suppose (d, i) is in the domain of B
◦
. Then, B
◦
σ[d,i],τ
: σ[d, i] τ.
iii. Suppose (d, i) B
◦
(d, j ). Then, B
◦
σ[d,i],τ[d, j]
: σ[d, i] ≡ τ[d, j].
We leave the easy proof to the reader. The following theorem shows that
we can decompose in and ≡ in two ways.
Theorem 3.1 We have:
i. σ τ if f, for some σ
, σ ≡ σ
and σ
τ.
ii. σ τ if f, for some τ
, σ τ
and τ
≡ τ.
Proof
Ad (i) Suppose B witnesses σ τ. Let Y be the range of B
◦
. Then, we have:
B
σ,τY
: σ ≡ τ Y and τ Y τ. Conversely, if σ ≡ σ
τ, then σ
σ
τ, and, hence, σ τ.
Ad (ii) We define τ
by:
• I
τ
,d
:= I
σ,d
∪ ({σ} × I
τ,d
).
• (d, ı, p) R
τ
(e, j, q) iff (ı ∈ I
σ,d
and j ∈ I
σ,e
and (d, ı, p) R
σ
(e, j, q)), or (ı = (σ, i) and j = (σ, j ) and (d, i, p) R
τ
(e, j, q)).
We note that I
σ,d
∩ ({σ} × I
τ,d
) is empty. Clearly σ τ
.
Suppose B witnesses σ τ. We define ı B
d
j iff (ı ∈ I
σ,d
and ı B
d
j),
or ı = (σ, j ). Clearly B
is a total and surjective bisimulation from τ
to τ, so τ
≡ τ.
3.2 Sachverhalte
We remind the reader that a Sachverhalt σ is defined as an atom of , that is:
σ is not ⊥⊥ and, whenever σ
σ, then σ
≡ ⊥⊥ or σ
≡ σ.
Theorem 3.2 We have: σ is an atom if and only if σ is bisimilar to a nonempty
connected sachlage.
Proof Suppose σ is an atom. Since σ is not the empty Sachlage, there is
an object occurrence (d, i) in σ. Clearly, σ[d, i] is nonempty and connected.
Moreover, σ[d, i] σ. Hence, σ ≡ σ[d, i].
For the converse it is sufficient to showthat a nonempty connected sachlage
σ is an atom. Suppose B: σ
σ and σ
is nonempty. Since the range of
B
◦
must be closed in σ, it must be equal to O
obj
σ
, so B is surjective. Hence
B : σ
≡ σ.
Theorem 3.3 Each Sachlage σ is the supremum of the Sachverhalte below it. In
other words: is atomistic.
Proof Suppose, for all Sachverhalte σ
0
σ, we have σ
0
τ. It follows that,
for each object occurrence (d, i) in σ, σ[d, i] τ; say this is witnessed by B
(d,i)
.
A Tractarian Universe
Each B
σ,τ
(d,i)
is a partial bisimulation between σ and τ. Clearly the union of the
B
σ,τ
(d,i)
will be a total bisimulation between σ and τ. Hence σ τ.
3.3 The SumOperation
We show that every set of Sachlagen has a supremum. Let S be a set of
Sachlagen. We define
S =: σ
+
as follows.
•
I
σ
+
,d
:= {(σ, i)  σ ∈ S and i ∈ I
σ,d
}.
•
(d, (σ, i), p) R
σ
+ (d
, (σ
, i
), p
) :⇔ σ = σ
and (d, i, p) R
σ
(d
, i
, p
).
We note that
is defined as an operation on sets of Sachlagen, where ≡ is
not divided out. We have:
Theorem 3.4 The operation
is the supremum operation for our ordering
(modulo ≡).
Proof Consider any σ in S. We define a bisimulation B from σ to σ
+
as
follows. Let i be in I
d
. We set: i B
d
(σ, i). It is easily seen that B is indeed a
total bisimulation. So, σ σ
+
.
Suppose B
σ
: σ τ, for each σ ∈ S. We define (σ, i) B
+
d
j :⇔ i B
σ,d
j. It is
easily seen that B
+
is a total bisimulation from σ
+
to τ. Hence σ
+
τ.
Note that we do not have enough information yet to construct the maximal
Sachlage as the sum of all Sachlagen: we do not yet know that we there
is a set of representatives that contains a representative for each Sachlage
(as an object modulo bisimulation). We will show that such a set exists in
Theorem 4.3.
We are mostly interested in properties of
modulo bisimulation, however
the following property is dependent on the precise implementation.
Theorem 3.5 Suppose S ⊆ S
. Then,
S
S
.
We leave the easy proof to the reader.
We say that a Sachlage σ is completely join prime, if, for every set of
Sachlagen S, if σ
S, then, for some σ
∈ S, σ σ
.
Theorem 3.6 Every Sachverhalt is completely join prime.
Proof Consider any Sachverhalt σ. Let S be a set of Sachverhalte. Suppose
B: σ
S. Let (d, i) be any object occurrence in σ and suppose i B
d
(σ
, i
).
Remember that σ[d, i] is a connected component of σ that is bisimilar to σ.
Let B
0
be a bisimulation from σ to σ[d, i]. We define B
between σ and σ
, by
j B
e
j
iff j B
0,e
j
0
and j
0
B
e
(σ
, j
), for some j
0
∈ I
σ[d,i]
. It is easy to see that
B
witnesses that σ σ
.
A. Visser
We give two definitions.
•
σ
is the set of σ[d, i] such that (d, i) is an object occurrence in σ. We could
also say:
σ
is the set of connected σ
0
such that σ
0
σ.
•
Let S and S
be sets of Sachlagen. We have: S ≡ S
iff, for all σ ∈ S, there
is a σ
∈ S
, such that σ ≡ σ
, and for all σ
∈ S
, there is a σ ∈ S, such that
σ ≡ σ
.
A Sachlage functions in many respects as a set of Sachverhalte.
Theorem 3.7 We have:
i. If S ≡ S
, then
S ≡
S
.
ii. σ ≡
σ
.
iii. For any Sachverhalt σ
0
with σ
0
σ, there is a σ
1
∈
σ
, such that σ
0
≡ σ
1
.
iv. σ ≡ σ
if f
σ
≡
σ
.
v. σ ≡ σ
if f, for all Sachverhalte σ
0
, σ
0
σ if f σ
0
σ
.
Proof (i) and (ii) are easy. For (iii), note that, since σ
0
σ ≡
σ
, we find,
that σ
0
≡ σ
1
, for some σ
1
in
σ
, since σ
0
is completely join prime.
We prove (iv). For the righttoleft direction, we use (i) and (ii). For the left
toright direction, suppose σ ≡ σ
and σ
0
∈
σ
. Then σ
0
σ
, so by (iii), there
is a σ
0
∈
σ
, such that σ
0
≡ σ
1
. Similarly, we prove that if σ
0
∈
σ
, then, for
some σ
0
∈
σ
, σ
0
≡ σ
0
. Hence
σ
≡
σ
.
We prove (v). From left to right is trivial. Suppose, for all Sachverhalte σ
0
,
σ
0
σ iff σ
0
σ
. Then, each element σ
0
of
σ
is below σ
. Ergo, σ ≡
σ
σ
. Similarly, σ
σ.
3.4 The ConjunctionOperation
We can define an operation on Sachlagen that produces an infimum in several
ways. We take one such way as our official choice. For some discussion of the
options, see Remark 3.10.
We define the conjunction σ τ of Sachlagen σ and τ as follows. Consider
the maximal bisimulation B between σ and τ. Note that B need not be total
or surjective; it might even be empty. Let X be the domain of B
◦
. We take
σ τ := σ X.
Theorem 3.8 σ
0
σ
1
is, modulo ≡, the inf imum of σ
0
and σ
1
.
Proof We will use our convention that for any bisimulation B
, the domain of
B
is the domain of B
◦
.
Let B be the maximal bisimulation from σ
0
to σ
1
.
Note that (σ
0
σ
1
) σ
0
, and hence, (σ
0
σ
1
) σ
0
. Moreover, B
στ,τ
: (σ
τ) τ.
Suppose C
i
: τ σ
i
, for i = 0, 1. We note that C :=
˘
C
0
; C
1
is a bisimulation
from σ
0
to σ
1
. The domain C of is the domain of
˘
C
0
, since C
1
is total. The
A Tractarian Universe
domain of
˘
C
0
is the range of C
0
. We may conclude that the range of C
0
is included in the domain of B, since B is maximal. Ergo: C
0,τ,σ
0
σ
1
: τ
(σ
0
σ
1
).
We prove distributivity of over
.
Theorem 3.9 We have σ
S ≡
{σ σ
 σ
∈ S}.
Proof Let σ
0
be a Sachverhalt. We have:
σ
0
_
σ
S
_
⇔ σ
0
σ and σ
0
S
⇔ σ
0
σ and, for some σ
∈ S, σ
0
σ
⇔ for some σ
∈ S, σ
0
σ and σ
0
σ
⇔ for some σ
∈ S, σ
0
(σ σ
)
⇔ σ
0
{σ σ
 σ
∈ S}
The desired conclusion follows from Theorem 3.7(v)
Remark 3.10 There are several alternative choices for our operation .
Perhaps the two most obvious are:
•
σ τ :=
{σ
σ  σ
τ}.
•
σ τ :=
{σ
σ  σ
is a Sachverhalt and σ
τ}.
Note that in both cases, we really do our summation over a set.
A third alternative is in in the spirit of our chosen definition of and
connects this to a beautiful observation: bisimulations can themselves be
considered as Sachlagen. Consider any bisimulation C between σ and τ. We
define a Sachlage (C) by:
•
I
(C),d
:= {(i, j )  i C
d
j},
•
(d, (i, j ), p) R
(C)
(e, (i
, j
), q) iff (d, i, p) R
σ
(e, i
, q).
Note that it follows that:
(d, (i, j ), p) R
(C)
(e, (i
, j
), q) iff (d, j, p) R
τ
(e, j
, q).
We can now easily show that (C) σ and (C) τ. Moreover, whenever
C
0
: ν σ and C
1
: ν τ, then (
˘
C
0
; C
1
) ≡ ν.
Clearly C
◦
⊆ C
◦
implies (C) (C
). This suggests the following
definition of . We define σ τ := (B), where B is the maximal bisimu
lation between the Sachlagen σ and τ.
3.5 The Sachlagen form a Set
We assumed that all the ingredients of what makes a logical space (D, T, M,
P, F) are just sets. It follows that any Sachlage can be modeled as a set.
A. Visser
However the totality of the Sachlagen is a proper class, since we have put no
constraint on the choice of the occurrences. Since, the sum of arbitrarily many
copies of a Sachlage is again a Sachlage, we would even obtain a proper class
if we could pick an element of every isomorphism type of the Sachlagen.
In Section 4, we will produce a unique normal form modulo bisimulation
for each Sachlage. As we will see these normal forms form a set. Hence the
Sachlagen form a set modulo bisimulation (Theorem 4.3). As a consequence
of this result, we find:
Theorem 3.11 In any logical space, we have a maximal Sachlage . The
Sachlagen modulo bisimulation ordered by form a atomistic distributive
complete lattice.
Remark 3.12 We can also define subtraction of Sachlagen, making the Sach
lage modulo bisimulation into a Boole Algebra.
We note that this Boole algebra is the Boole algebra of Sachlagen in their
inclusive role. It is not the Boole Algebra that Wittgenstein defined over logical
space. Wittgenstein’s algebra is defined on sets of Sachlagen in their exclusive
role. A Sachlage σ in the inclusive role reappears as a set of Sachlagen in their
exclusive role as ↑σ := {τ  σ τ}.
11
Our sum corresponds to Wittgenstein’s
conjunction: ↑
{σ, τ} := ↑σ ∩ ↑τ.
4 Normal Forms for Sachlagen
In this section, we provide three kinds of normal form for Sachlagen. The
last two kinds make a bisimulationfree approach to Sachlagen possible. The
first two kinds of normal forms are the usual normal forms familiar from e.g.
modal logic: bisimulation collapse and canonical unraveling. The last kind,
canonical bisimulation collapse, consists of the bisimulation collapses of the
canonical unravelings. Since, our Sachlagen are more constrained than most
kinds of transition systems the normal forms are, in a sense, better. E.g.,
for canonical unraveling in modal logic one needs to choose a cardinal. This
is unnecessary when one studies graded modalities and employs counting
bisimulation. Sachlagen and bisimulation on Sachlagen are more like graded
transition systems and counting bisimulation. (See e.g. [3]).
4.1 Bisimulation Collapse
We can find a minimal representation of a Sachlage σ := (I, R) as follows. Let
B be the set of all bisimulations between σ and itself. This set is not empty,
11
I am ignoring the fact that we do not have a set here. We clearly can tell a story, e.g. using the
normal forms of the next section, to make sense of this.
A Tractarian Universe
since id
σ
is in it. Let B
†
:=
B. So B
†
is the maximal bisimulation of σ with
itself. By Theorem 2.3, each B
†
d
is an equivalence relation. We write [i]
B
†
d
for
the B
†
d
equivalence class of i in I
d
. We define coll(σ) := σ
†
by:
•
I
†
d
:= I
d
/B
†
d
,
•
(d, [i]
B
†
d
, p) R
†
(e, [ j]
B
†
e
, q) :⇔ ∃i
0
∈[i]
B
†
d
, j
0
∈[ j]
B
†
e
(d, i
0
, p) R (e, j
0
, q).
It is easy to see that σ
†
is indeed a Sachlage and that σ
†
≡ σ. Moreover, we
find that, whenever σ
0
≡ σ
1
, then σ
†
0
and σ
†
1
are isomorphic.
In the next subsection, we will develop normal forms using canonical
unraveling. These normal forms are unique simpliciter. As soon as we have
these, the bisimulation collapse of such a canonical unraveling is a normal form
that is both bisimulationminimal and uniquely determined for each Sachlage.
4.2 The Canonical Unraveling
Consider a logical space. We consider (nonempty) sequences f of the form
(d
0
, p
01
, p
10
, d
1
, p
11
, p
20
, d
2
, . . . , p
n0
, d
n
).
Here p
01
is in P
d
0
, and, for k > 0, p
k0
and p
k1
are in P
d
k
and p
k0
= p
k1
.
Moreover we demand: F
d
k
( p
k1
) M F
d
k+1
( p
(k+1)0)
. We call these sequences
paths. The length of a path is the unique n such that the length of the path
qua sequence is 3n +1. So, a path of length 0 has the form (d).
Since the objects form a set and the poles of an object form a set, the paths
form a set.
We define:
•
f # g iff, for some d, both f and g start from d, and either f = (d) or g =
(d) or ( f = (d, p, . . .) and g = (d, p
, . . .) and p = p
).
•
˘
f is the result of reading f backwards.
•
f g is defined if
˘
f # g. If our operation is defined, we have:
f
, .. ,
(. . . , d)
g
, .. ,
(d, . . .) = (. . . , d, . . .).
We note that the result is again a path.
The paths with form a partial monoid with as units the paths (d). Warning:
the paths do not generally form a category. E.g., (d, p, e) (e) (e, p, d) is not
defined, even if both (d, p, e) (e) and (e) (e, p, d) are.
A set U of paths is coherent iff, for some object d,
i. U contains the path (d).
ii. All elements of U have starting point (domain) d.
A. Visser
iii. For any f and g in U, there are h, f
0
, g
0
, such that f = h f
0
, g = h
g
0
and f
0
# g
0
. In other words, either one of the paths (weakly) extends
the other, or, after sharing an initial part they diverge by taking different
poles.
12
iv. U is closed under initial subpaths.
v. Suppose f = (. . . , q, e) is in U and r is a pole of e unequal to q. Then,
there are r
and e
, such that f (e, r, r
, e
) is in U. In other words, when a
path can be extended, it is extended in U.
We note that, the totality of the coherent sets of a given space forms a set.
A coherent set is something like a map that a fairy living in a Sachlage would
make of the Sachlage, assuming that the fairy would assume the Sachlage to be
stable, in the sense that backtracking leads to the same places, and that the fairy
would never unnecessarily identify places.
Consider any coherent set U. We associate a Sachverhalt svh(U) := σ
U
to
U in the following manner. We define:
•
I
σ
U
,e
:= { f ∈U  cod( f ) = e},
•
(e, f, q) R
σ
U
(e
, f
, q
) iff f = f
(e
, q
, q, e) or f
= f (e, q, q
, e
).
We easily see that σ
U
is a Sachverhalt. Since the coherent sets form a set,
also the σ
U
form a set.
Theorem 4.1 Given any Sachlage σ and any object occurrence (d, i) in σ, we
can f ind a coherent set cs(σ, (d, i)) := U
σ,(d,i)
, and a functional bisimulation G
(or, more explicitly, G
σ,(d,i)
) from svh(cs(σ, (d, i))) to σ, such that G((d, (d)) =
(d, i). Moreover, for any bisimulation B from svh(cs(σ, (d, i))) to σ with
(d, (d)) B
◦
(d, i), we have G
◦
⊆B
◦
.
Proof We define U
σ,(d,i)
as follows. By recursion on path length we simul
taneously define elements of U and a mapping G
0
on these elements such
that G
0
( f ) ∈ I
cod( f )
. We put (d) in U and set G
0
((d)) := i. We treat the
case that f = (. . . , q, e) and G
0
( f ) =: j are already given. The case starting
from (d) is similar. Suppose q
∈ I
e
\ {q}. Then, for some unique j
, q
, e
,
we have (e, j, q
) R
σ
(e
, j
, q
). We add f
:= f (e, q
, q
, e
) to U, and take
G
0
( f
) := j
.
We take (e, f ) G (e, G
0
( f )), for e = cod( f ). It is not difficult to see that G
is a functional bisimulation from svh(cs(σ, (d, i))) to σ. That G
◦
is part of any
bisimulation from svh(cs(σ, (d, i))) to σ that connects (d, (d)) to (d, i) follows
with an easy induction on path length.
12
We note that this could also be formulated as follows: consider two different paths f and g in U.
Let k be the smallest number such that either (( f )
k
is defined and (g)
k
is undefined) or (( f )
k
is
undefined and (g)
k
is defined) or (( f )
k
and (g)
k
are both defined and different), then k is 3k
0
+1,
for some k
0
(assuming that we start counting with 0).
A Tractarian Universe
Theorem 4.2 Suppose σ is a Sachverhalt and (d, i) is an objectoccurrence in σ,
then σ ≡ svh(cs(σ, (d, i))).
Proof Since, G
σ,(d,i)
is surjective to σ[d, i] and since σ is a Sachverhalt, we find
σ ≡ svh(cs(σ, (d, i))).
Thus, we see that all Sachverhalte are bisimilar to σ
U
, for some coherent U.
We are now in the position to give our canonical unraveling of σ. This uses, of
course, that the σ
U
form a set.
•
unr(σ) :=
{σ
U
 U is coherent and σ
U
σ}.
Theorem 4.3 We f ind:
a. For each Sachlage σ, unr(σ) ≡ σ.
b. If σ ≡ τ, then unr(σ) = unr(τ).
c. If σ τ, then unr(σ) unr(τ).
d. The canonical unravelings unr(σ) form a set.
Proof Our ordering is atomistic and every atom is bisimilar to a σ
U
. Hence,
unr(σ) ≡ σ. This gives us (a). Moreover, (b) is trivial, and (c) is immediate
from Theorem 3.5. Finally, since the σ
U
form a set, say X, the subsets of X
form a set. Hence, the objects of the form unr(σ) form a set.
We note that our earlier result that the Sachlagen modulo bisimulation
form a distributive atomistic complete lattice follows immediately from Theo
rem 4.3, since is essentially the subset relation.
The material we developed up to this point is sufficient for the purposes
of our normal form theorem. However, there is a bit more to say about our
construction. We have the following theorem.
Theorem 4.4 Let U be a coherent set starting with (d). Suppose B is a bisimula
tion fromσ
U
to σ with (d, (d)) B
◦
(d, i). Then U = cs(σ, (d, i)) and G
◦
σ,(d,i)
⊆B
◦
.
Proof Let G := G
σ,(d,i)
and let G
0
be the corresponding function of paths. We
show by induction on path length that W := cs(σ, (d, i)) is a subset of U and
f B G
0
( f ).
Clearly, (d) is in U and (d) B i.
Suppose (d, p, q, e) is in W and G
0
((d, p, q, e)) = j. Then, we have
(d, i, p)P
σ
(e, j, q). Since (d, (d)) B
◦
(d, i), we find that (d, (d), p) R
σ
U
(e, h, q)
and h B j, for some h. The only possible such h is (d, p, q, e). So (d, p, q, e) is
in U and (d, p, q, e) B j.
Suppose we have the induction hypothesis for (. . . , q, e). Let G((. . . , q, e)) =
j. Suppose (. . . , q, e, p
, q
, e
) is in W and G
0
((. . . , q, e, p
, q
, e
)) = j
. Since,
(e, (. . . , q, e)) B
◦
(e, j ), we find that (e, (. . . , q, e)) R
σ
U
(e
, h
, p
) and h
B j
,
A. Visser
for some h
. Since p
= q, h
must be an extension of (. . . , q, e). Ergo, h
=
(. . . , q, e, p
, q
, e
). So, (. . . , q, e, p
, q
, e
) is in U and (. . . , q, e, p
, q
, e
) B j
.
We may conclude that W ⊆ U. It is immediate that W = U.
As an immediate corollary we have:
Theorem 4.5 If U is a coherent set starting with (d), then U =
cs(svh(U), (d, (d))).
Remark 4.6 The reader who has some knowledge of category theory, will
suspect there is an adjunction here. That is indeed the case. Consider the
category with as objects the pointed Sachlagen (σ, (d, i)) and as morphisms
between (σ, (d, i)) and (τ, (e, j )) the bisimulations B (not necessarily total or
surjective) such that (d, i) B
◦
(e, j ) (and, hence, d = e). Consider as second
category the coherent sets with as morphisms just the identity. Then svh
+
is
the left adjoint of cs, where svh
+
(U) := (svh(U), (d, (d)) for (d) the starting
point of U.
Our normal forms still ‘contain’ a number of bisimilar copies of a
Sachverhalt, since σ
U
≡ σ
V
, does not imply σ
U
= σ
V
. We show that the situa
tion is a bit better: σ
U
≡ σ
V
, does imply σ
U
is isomorphic to σ
V
. So, for each
Sachverhalt, our normal form contains only isomorphic copies.
Theorem 4.7 Suppose σ
U
≡ σ
V
. Then, σ
U
is isomorphic to σ
V
.
Proof Suppose B witnesses that σ
U
is bisimilar to σ
V
. Suppose U starts with
(d) and (d) B h. By Theorem 4.4, we find that U = cs(σ
V
, (d, h)) and G :=
G
σ
V
,(d,h)
is a functional subbisimulation of B.
Suppose that V starts with (e). It is easily seen that
˘
h G (e, (e)). Applying
Theorem 4.4 to σ
V
and
˘
G, we find that V = cs(U, (e,
˘
h)). Let G
:= G
σ
V
,(e,
˘
h)
.
We note that G; G
is a functional bisimulation from σ
U
to σ
U
. Moreover,
(d) (G; G
) (d). Applying Theorem 4.4 for a third time, we find that G; G
is
the identity bisimulation of σ
U
. Similarly, G
; G is the identity bisimulation of
σ
V
. We may conclude that σ
U
and σ
V
are isomorphic.
4.3 The Canonical Collapse
The canonical collapse is obtained by collapsing the canonical unraveling.
Thus, we define the canonical collapse of σ by:
•
ccoll(σ) := coll(unr(σ)).
This collapse is evidently bisimulation minimal. Moreover, we have: if σ ≡ τ,
then ccoll(σ) = ccoll(τ). So each ≡equivalence class has a unique normal
form. All this is trivial by the preceding results on coll and unr. However, we
also have the following theorem.
A Tractarian Universe
Theorem 4.8 Suppose σ τ. Then, ccoll(σ) ccoll(τ).
Proof Suppose σ τ. Let B be the maximal bisimulation on unr(σ) and let C
be the maximal bisimulation on unr(τ).
Consider ı in I
ccoll(σ),d
. The occurrence ı is of the form [(σ
U
, f )]
B
d
, where
(d, (σ
U
, f )) is an objectoccurrence in unr(σ), and, hence, σ
U
σ.
Clearly, (d, (σ
U
, f )) is also an objectoccurrence in unr(τ). It follows that
the occurrence [(σ
U
, f )]
C
d
is in I
ccoll(τ),d
. So it is sufficient to show that
[(σ
U
, f )]
C
d
= [(σ
U
, f )]
B
d
.
Suppose (σ
U
, f ) C
d
(σ
V
, g). It follows, by Lemma 3.1(iii), that the
Sachverhalte ¯σ
U
:= (unr(τ))[(d, (σ
U
, f ))] and ¯σ
V
:= (unr(τ))[(d, (σ
V
, g))] are
bisimilar. But, ¯σ
U
is an isomorphic copy of σ
U
and ¯σ
V
is an isomorphic copy of
σ
V
. Ergo, σ
U
is bisimilar to σ
V
, and, hence, σ
V
σ. It follows that (d, (σ
V
, g)) is
an objectoccurrence in unr(σ). Since, (d, (σ
U
, f )) and (d, (σ
V
, g)) are bisimilar
according to C
◦
restricted to unr(σ) and since B is maximal on unr(σ), we
find (d, (σ
U
, f )) B
◦
(d, (σ
V
, g)). We may conclude that (σ
V
, g) ∈ [(σ
U
, f )]
B
d
.
Hence, [(σ
U
, f )]
C
d
= [(σ
U
, f )]
B
d
, as desired.
Thus the order structure takes a very simple form on canonical collapse
normal forms.
5 Perspectives
In this section, we briefly touch on possible variations, improvements and
extensions of the framework.
5.1 Configurations
What kind of structures do we get when we drop the demand of saturatedness?
Let’s call these structures conf igurations. A conf iguration γ is defined as
follows.
•
A function I from objects to (possibly empty) sets of indices.
•
A binary relation R on O
pole
γ
, where O
pole
γ
, the set of pole occurrences, is
the set of (d, i, p), where i is in I
d
and p is in P
d
. We demand that R is
functional and symmetrical and that (d, i, p) R (d
, i
, p
) implies F
d
( p) M
F
d
( p
).
So a configuration is a Sachlage minus totality. We define:
•
P
γ
:= {(d, i, p)∈O
pole
γ
 (d, i, p) ∈ dom(R
γ
)}.
•
F
γ
(d, i, p) := F
d
( p).
Note that with every object d there corresponds a configuration [[d]] given by
(i) I
d
:= ∅, if d
= d, and I
d
:= {0}, if d
= d, and (ii) R is the empty.
A. Visser
We define a simulation B from γ to γ
just as we defined a bisimulation
before, now only asking for the forward property. We define a bisimulation as
before. We define γ γ
iff there is a total simulation from γ to γ
. We count
γ and γ
the same iff there is a total and surjective bisimulation between them.
Note that, on Sachlagen our new coincides with our old . However, on
Sachlagen the induced equivalence relation of coincided with the presence of
a total and surjective bisimulation. This is not true anymore for configurations.
We note that we can define the existence of a configuration γ as the fact
that γ σ, for the maximal existing Sachlage σ. We can say that an object d
exists if [[d]] does. Thus, the fixed store of objects in the tractarian universe
can be best conceived of as a store of possible objects, that come into existence
if they occur in some fact.
Associated with configurations, there is a lingering doubt: is the framework
we developed a ref lective equilibrium? There is a feeling that configurations
are something like complex objects. E.g., when we think about a morphism
between logical spaces, we could imagine mapping objects to configurations.
However, the poles of a configuration γ do not quite behave as the poles of
objects. If we multiply the poles of an object it is always in a coordinated
way: we copy the whole set. However, if we take a bisimulation variant
of a configuration, it is clear that some groups of poles will be linked in
multiplication or contraction, but some groups are not. Can we generalize the
notion of object in such a way that objects and configurations behave in the
same way?
5.2 Argument Places
As we already noted, we implement the most straightforward idea of what the
places are. It is clear that we can build in refinements of this treatment rather
easily. E.g., on each set of poles P
d
we can have a group of permutations G
d
.
Perhaps we can improve the notion of bisimulation by taking B as a function
on objects that assigns to each d a ternary relation B
d
between I
d
and G
d
and
I
d
, such that B
d
satisfies the zig and the zagproperty:
•
If i
0
B
g
0
d
0
i
0
and (d
0
, i
0
, p
0
) R (d
1
, i
1
, p
1
), then, for some i
1
in I
d
1
and g
1
in
G
d
1
, we have i
1
B
g
1
d
1
i
1
and (d
0
, i
0
, g
0
p
0
) R
(d
1
, i
1
, g
1
p
1
) (zigproperty),
•
If i
0
B
g
0
d
0
i
0
and (d
0
, i
0
, p
0
) R
(d
1
, i
1
, p
1
), then, for some i
1
in I
d
1
and g
1
in
G
d
1
, we have i
1
B
g
1
d
1
i
1
and (d
0
, i
0
, g
−1
0
p
0
) R (d
1
, i
1
, g
−1
1
p
1
) (zagproperty).
I do not know whether this works. Also, I do not think that we can treat all
examples produced by Kit Fine in this way.
5.3 Space and Time
How do time and space fit into a Wittgensteinean universe? One option would
be to take them to be external to the Sachverhalte. However, this has several
problems. First it is hard to see how it coheres with ‘the world being all that
A Tractarian Universe
is the case’. It is hard to swallow that temporal and spatial relations would be
outside the world, part of the unspeakable, so to say. Secondly, if we refuse
to put constraints on how subsequent snapshots of the world cohere, then it
difficult to understand how one could seriously speak of time at all. There’s
just nothing connecting the moments.
If we put spatiotemporal locations as objects inside the Sachlage, this gives
us the usual problems about how to model, e.g., temporal order. Moreover,
consider John is angry. It is plausible that John will have a temporal
location argument place and that angry will carry one too too. But inside the
Sachverhalt we would like these places to be filled with the same location. The
present framework does not seem to have means to enforce this idea. Again, it
would be good to find an appropriate enrichment.
Acknowledgements I thank Jaap van der Does, Leon Geerdink, Jesse Mulder, Vincent van
Oostrom, Martin Stokhof, Göran Sundholm, and the members of the Utrecht Tractatus Reading
Group for their comments and criticisms. I am grateful to the anonymous referee for his/her
helpful suggestions. For an acknowledgement of priority to Alan Bawden, see Section 1.6.
Open Access This article is distributed under the terms of the Creative Commons Attribution
Noncommercial License which permits any noncommercial use, distribution, and reproduction in
any medium, provided the original author(s) and source are credited.
References
1. Aczel, P. (1988). Nonwellfounded sets. CSLI lecture notes (Vol. 14). Stanford: CSLI.
2. Black, M. (1971). A companion to Wittgenstein’s Tractatus. Cambridge: Cambridge University
Press.
3. D’Agostino, G., & Visser, A. (2002). Finality regained: A coalgebraic study of Scottsets and
multisets. Archive for Mathematical Logic, 41, 267–298.
4. Fine, K. (2000). Neutral relations. The Philosophical Review, 109, 1–33.
5. Leo, J. (2008). The identitity of argumentplaces. The Review of Symbolic Logic, 1(3), 335–354.
6. Leo, J. (2008). Modeling relations. Journal of Philosophical Logic, 37, 353–385.
7. Marion, M. (2004). Ludwig Wittgenstein, Introduction au “Tractatus logicophilosophicus”.
Philosophies. Presses Universitaires de France, Paris.
8. Rieger, A. (2000). An argument for Finsler–Aczel set theory. Mind, 109(434), 241–253.
9. Wittgenstein, L. (1961). Tractatus logicophilosophicus. London: Routledge & Kegan Paul.
A. Visser
together to form saturated Sachlagen and Sachverhalte.1 The Sachverhalte are atomic Sachlagen—in a sense of atomic that will be explicated. An important point of the paper is the analysis of sameness of Sachlagen. This is an issue that Wittgenstein did not address, probably because he never attempted to build a concrete model of his Logical Space. As will be explained in the paper we opt for bisimularity as the appropriate analysis of sameness. This means, very roughly, that two Sachlagen are the same as they cannot be distinguished from an internal structural point of view. 1.1 Saturated Versus Unsaturated Frege insisted that only in the context of a sentence does a word have meaning (the context principle). Now consider the sentence Roel is happy. We can view the predicate expression is happy as obtained by removing the name Roel from the sentence. So the predicate expression really is: ( ) is happy. We can now see that the predicate expression is ergänzungsbedürftig or unsaturated: it needs supplementation to form an expression that is meaningful in its own right, to wit the sentence. This Ergänzungsbedürftigkeit is, in Frege’s set up, inherited on the semantic level by the predicate itself. But what about referring expressions? What about Roel? It seems to me that if we were really serious about the context principle, we should say that Roel was really Roel ( ), and consequently that the object corresponding to the name is unsaturated. If we forget philosophy of language for a moment, we could say that, in this view, properties and relations can only have reality in the objects having them, but, conversely, that objects also only have reality in the properties they have and in the relations they stand in. Such a view would blur the distinction between object and property/relation. The tradition has refrained from taking the step towards unsaturated objects. E.g., in Montague Grammar, we have a basic type of objects and a basic type of propositions/truth values. All other types are derived from these. The objects of the basic types can be viewed as the saturated entities in this framework: these are the entities that are not functions. Wittgenstein, in my reconstruction, insisted that objects are ergänzungsbedürftig too. This is reflected in the fact that they do not occur independently. Here is a telling quote:2 2.0131 Der räumliche Gegenstand muß im unendlichen Raume liegen. (Der Raumpunkt ist eine Argumentstelle.) Der Fleck im Gesichtsfeld muß zwar nicht rot sein, aber eine Farbe muß er haben: er hat sozusagen den Farbenraum um sich. Der Ton muß eine Höhe haben, der Gegenstand des Tastsinnes eine Härte, usw.
will leave “Sachlage” and “Sachverhalt” untranslated. They are Tractarian expressions with very special meanings. Still, the Pearce & McGuiness translation of “Sachlage” as situation, and of “Sachverhalt” as state of af fairs, does reflect my interpretation of these words reasonably well. 2 We use the Pearce & McGuiness translation [9] throughout the paper.
1I
Remark 1.21+2. in his classical [2]. 3. We note that Black is discussing the unsaturatedness of names. without intermediaries. objects of the sense of touch some degree of hardness. ‘unsaturated’.3 There are no relations and traditional objects. must have some colour: it is.A Tractarian Universe 2. and so on. Tractatus 3. The former being ‘incomplete’. p 13. the negative answer nominalism. and their esse is to be eligible for such concatenation (3.0123 which is about objects. ‘in need of supplementation’. (A spatial point is an argumentplace. They simply need not have traditional counterparts. He says: Frege took a decisive step forward when he introduces a radical distinction between ‘functions’ and ‘objects’. like links of a chain (cf. Do they occur as objects in configurations. We take the unsaturatedness of names to be reflected in the unsaturatedness of objects. .3). In short.) The above discussion takes our ordinary understanding of world and language as starting point and reasons from there about what the tractarian objects could be. If we start from the standpoint of Wittgenstein’s ontology. have to be symbolized by words of a peculiar and distinctive sort (‘function names’). our leading idea of a rapprochement between objectsquaparticulars and universals because they are all just unsaturated. Notes must have some pitch.0123). if you wish. here is the discussion by Max Black. Names occur only in association with other names. the unsaturatedness of names to reflect the unsaturedness of objects.1432 seems to speak against the objectual understanding of universals. not of objects. though it need not be red.14). 3I thank Jesse Mulder for stressing this point to me. happiness cannot occur without someone being happy. Since Tractarian objects are not given to us in ‘surface language’. but Roel cannot occur without being in some mood. it might be said. or do they subsist in the way objects click together? The positive answer to this question can be labeled realism. and have no reference except in the propositional context (3. strongly suggests an equal treatment of a and R. having ‘gaps’ that need top be filled by names of objects. all simple symbols (names) were unsaturated in something like Frege’s sense. we just have unsaturated objects.) A speck in the visual field. in elementary propositions they grip one another.1 Nominalism versus Realism: There has been some discussion whether universals like R are objects in Wittgenstein’s sense. so to speak surrounded by colourspace. we could easily imagine that for some of these objects it is impossible to say whether they are more like traditional objects or traditional relations.203+3. On the other hand.0131 A spatial object must be situated in infinite space. For example. (Black refers to 2. or. Wittgenstein went still further: for him. This reading finds some support in the exegetic literature.
Wittgenstein’s discussion of identity could carry some suggestion that he would not allow configurations of this form. See for a discussion of exegetical matters concerning the nominalism versus realism issue. it is neutral between realism versus nominalism.g. The notion of logical form does not necessarily imply a substitution principle. More generally an object that has n poles that can click together both with themselves and with each other. If in the Sachverhalte nothing corresponds with the Pi . If they were setlike. the chapter on Problèmes ontologiques in [7]. . . we could suppose that a has two poles p. then we can always substitute a for b in a configuration. we can raise the number of possible properties of a by working with more poles that fit together. For example. the motivation of our framework. can form a oneobject Sachverhalt in n! ways. So there are precisely n! unary predicates applicable to a. but it certainly does cohere well with a substitution principle. Since our framework just models the ontology. All this seems rather unattractive. Visser We have already stressed that the question concerning traditional objects and relations. fits better with the realist option.4 This would give us two possible Sachverhalte only involving one occurrence of a: one where each pole clicks with itself en one where the poles click with each other. say p and q. looking from language to ontology.. e. . so that more interesting patterns of clicking together become possible.A. there is a powerful reason not to disallow such configurations.g.2 Objects in Multiplicity How precisely do the objects form a chain or configuration? We will assume that configurations are not setlike. where the Pi are supposed to be different. But then consider [Pa]. It is more clumsy to interpret familiar patterns of predication nominalistically.3. However. Saturatedness of [aRa] dictates that a does not have more poles. Thus. Thus [aRa] can be ‘derived’ from [aRb ]. However.. This will be discussed in more detail in Section 1. Here a can have only one pole. Sachverhalte [P0 a]. where both poles may click together both with themselves and with each other. does not arise when we start from the ontology. [P1 a]. pp. The pole p combines with the first pole of R. Should we 4 We say that the objects are connected via poles. how could we explain the difference between [aRb ] and [b Ra]? They are more like molecules where the objects are connected via links or poles. of the traditional subjectpredicate form. We would like to demand the following: if a and b have the same logical form. Consider a Sachverhalt of the form [aRa]. 68–84. the differences between the Sachverhalte must be accounted for as different ways in which a is linked to itself. Of course. Consider. we can make the different poles of a of different types. q. . e. How can we think about [aRa]? One idea is that a has two poles. 1. It feels like a hack. to wit intersubstitutivity of objects of the same logical form in Sachverhalte. . the pole q combines with the second pole of R.
We need not stipulate a varying number of poles to account for the possibility of both [aRa] and [Pa]: we can say that a has two occurrences in [aRa]. This would give us ‘polyadic objects’. how can this pole be employed twice? This phenomenon can be explained by the insight that what we find in Sachverhalte are occurrences of objects.012 In logic nothing is accidental: if a thing can occur in a state of affairs.3 How Objects Click Together The poles are what connects objects to each other. The reason for this choice is the intuition that the neatest way 5 Varying assemblies of poles are suggested as one possible option for obtaining a more systematic view of objects as special cases of (possibly) unsaturated configurations. As we will see. Thus. in 2. Each pole connects to another pole.A Tractarian Universe allow a to have a varying number of poles? But if this number varies. in our framework. See Section 5. Also. it should vary when we move from [aRa] to [Pa]. even their cardinality is indeterminate. The framework can easily be adapted to multilinking.6.6 Again this is just an assumption for the sake of simplicity. They are not indiscernibles. in many examples. for further remarks on this matter. at first sight. Since they are supposed to be saturated. We see no knockdown philosophical argument against the possibility of a varying number of poles for some objects or relations. a has exactly one pole. However. Not all poles may click together. but. so what varies here to make a ‘change’ its number of poles? We consider a variant of our example: presumably both [aRb ] and [Pa] could be Sachverhalte. Not only are they like indiscernibles. 6 This assumption makes our Sachlagen into linear graphs.5 We will also assume that connection is always a matter of two poles. we will assume that our poles have individuality. we will model our logical spaces by implementing a notion of occurrence of an object. the possibility of the state of affairs must be written into the thing itself. for a reference. This is assumed just for the sake of simplicity. in each case. it seems pretty straightforward to generalize our framework to accommodate variable sets of poles.012. We will assume that the poles of an object are fixed. Which types match is given with the specification of logical space. Each pole has a type. 2. We note there is some textual evidence that Wittgenstein thought about objects as occurring in states of affairs. . Ony poles with types that match may connect. Finally. he says: 2. occurrences have very thin identities. See Section 1. For example. 1.012 In der Logik ist nichts zufällig: Wenn das Ding im Sachverhalt vorkommen kann so muß die Möglichkeit des Sachverhaltes im Ding bereits präjudiziert sein.
Thus.4 Sachlagen and Sachverhalte Sachlagen are like molecules. they are not coded in the types. Visser to analyse a firstorder style predication like [Pab ]. pictures are always full pictures of possible worlds. The actual world is the existing Sachlage in its exclusive role. Maybe the existing Sachverhalte are in black and blue and red and the nonexisting ones in yellow and pink and sepia? Such a subtle convention. but as we will see it is fully determined by a set of Sachverhalte. das Bestehen und Nichtbestehen von Sachverhalten. Wittgenstein takes the exclusive reading as the default. and a Sachlage in the exclusive role corresponds to a conjunction of atomic sentences and negated atomic sentences (together: atomic literals) in which all possible atomic sentences occur precisely once.. This is somewhat mysterious. which sits on top of the picture theory. vor. In other words. by giving P two poles of type. Its interpretation is handled by the truthfunctional semantics. It is neutral with respect to existence. Sachlagen are the primary candidates for existence. According to our interpretation. in their exclusive role. 2. A Sachlage in the in the inclusive role corresponds to a conjunction of atomic sentences. because Wittgenstein includes the nonexistence of Sachverhalten as something that is depicted. there are no disjunctive pictures. So the argument places are given by the individuality of the poles. In their inclusive role. In Example 2. say. Again it is easy to adapt the framework to have indiscernible poles. a and a and b each one pole of type. they are full possible worlds.11 A picture presents a situation in logical space. b. this idea is worked out. This . We can take them as presenting something of the form at least this is the case (inclusive role) or as presenting this is the case and it is all that is the case (exclusive role). There is an important interpretive choice in our treatment of Sachlagen.11: 2. So a sentence of the form p ∨ q. the poles associated to an object would be a multiset of types. A Sachlage is not a set of Sachverhalte. Here a and b match. if all poles were indiscernible. 1.11 Das Bild stellt die Sachlage im Logischen Raume.g. They are saturated configurations of objects which are clicked together via their poles in the ways that are allowed by the nature of logical space.2 We remind the reader of what Wittgenstein says in 2. say. where p and q are atomic does not depict. the existence and nonexistence of states of affairs. Sachlagen appear in two roles. In the framework they are the smallest nonempty Sachlagen. surely would be beyond the range of simple depiction? In an alternative reading. what is depicted would always be the Sachlage in its exclusive role.7. Sachlagen are parts of possible worlds and. An existing Sachlage is a Tatsache or fact. Sachverhalte are special Sachlagen. A Sachlage is a (possible) situation that is depicted in a picture. Remark 1.A. E.
but there is no claim that the items constructed are Sachverhalte and Sachlagen. Suppose a and b match. The objects we provide to represent basic items like Sachverhalte and Sachlagen are supposed to represent Sachverhalte and Sachlagen as they are philosophically intended. respectively. Are all these possibilities the same or different? 7 Of course. Consider.. Remark 1. This shows that it is conceivable that at least some objects are. In fact. it seems improbable that even Wittgenstein would have such a subtle reading of what a picture does. etcetera. at least.7 For more on configurations. 1.A Tractarian Universe reading would solve how nonexistence can be pictured. we could hold that an object is conceptually different from the configuration containing just that object. I think that is the best way of looking at it. with urelements for items like objects and poles. coincide with Sachlagen.5 When Two Sachlagen are the Same In the present paper. We could also have two occurrences of a. however. We will prove the theorem that Wittgenstein used but failed to state: the subSachlage ordering is atomistic. without commenting upon it. see Section 5. Moreover. It would fit my interpretation better if Wittgenstein had omitted the phrase about nonexistence. Between Sachlagen (in the inclusive role) we have the subSachlage partial ordering. What is more the objects we construct contain artifacts of the modeling. A Sachverhalt is an atom in this ordering.3 The framework allows that there are saturated objects. i. say. every Sachlage is the supremum of the Sachverhalte below it. This means that a Sachverhalt is a minimal nonempty Sachlage.e. an object a with two poles p and q of type. aspects that are not present in the Sachverhalte and Sachlagen as philosophically intended. we could have one occurrence of a that links with itself. In this case. As a modeling medium we employ informal set theory. a and b. or. that the poles of an object form a set. such that each links with the other. the conf igurations. Also both Sachlagen and objects belong to an encompassing category. We are assuming that the urelements form a set. A Sachlage will simply be a saturated configuration of objects. One important matter is the question of sameness of two (models of) Sachverhalte. it diverges substantially from the ordinary understanding of picture. The objects that we produce in our modeling to represent Sachlagen are neutral with respect to inclusive or exclusive role. we are constructing models of logical spaces against the background of a classical metatheory or modeling medium. for example. We could have an infinity of occurrences of a that link with the order type of the integers. . Or we could have three occurrences of a that link in a circle. say.
i) and (e. see also Adam Rieger’s eminently readable paper [8]. j) in τ . We may compare this idea with the intrinsic notion of curvature in a geometry that can be understood independently of an embedding space. Of course. Consider two Sachlagen σ and τ . and representations of Sachverhalte “Sachverhalte”. Visser My intuition about this question can best be explained by introducing the idea of the inner perspective on a Sachlage. She would have no access to the identity of the occurrence: the occurrence is ‘thin’. say “pre 8 For a discussion of individuation of sets. The application of bisimulation that is. . which is impossible—even for a fairy. It is my feeling that the alternative of using different names. What would that be like? The fairy would be at an occurrence of an object and would be able to travel across the poles to other occurrences. We refer the reader to Peter Aczel’s wonderful book [1]. i) in σ and any objectoccurrence (e. j). the link across which she travels. It is also employed in Modal Logic to provide a notion of sameness of Kripke models. for every objectoccurrence (d. We say that (d. carefully noting along which poles they travel. Our intended notion of sameness can be technically explicated using the well established notion of bisimilarity. if there is no way that Femke and Inge can find a difference between them by moving from object occurrence to object occurrence.A.8 Remark 1. the Rieger paper does not have so much emphasis on notions of bisimulation. Suppose we drop fairy Femke in (d. If they cannot possibly find differences in this process. perhaps. indistinguishability implies that d = e. So. My idea is that two Sachlagen are the same if they are the same from the inner perspective. The Sachlagen σ and τ are the same if. there is an internally indistinguishable objectoccurrence (d. It and related concepts provide a good tool for proving various theorems. the initial occurrences are internally indistinguishable. and vice versa. Leaving breadcrumbs would change the Sachlage. She would be able to identify those things that have true individuality: the object where she currently is. i) and fairy Inge in (e. in Process Algebra it is used to explicate sameness of processes. j ) in τ . or even stronger: there is nothing to the Sachlage than the inner perspective. j) are internally indistinguishable or observationally equivalent. We note that Femke and Inge are not allowed to leave breadcrumbs. i) in σ . closest to the use we are making of it in this paper. The outer perspective only exists as an artifact of the modeling. Inge and Femke try out paths through their respective Sachlagen and check whether they can find differences. in that we call representations of Sachlagen “Sachlagen”.4 There is some abus de langage. Suppose we had a fairy living in a Sachlage. is the analysis of nonwellfounded sets. Consider any objectoccurrence (d. it is just a ‘multiplicity’. Unfortunately. We suppose that Femke and Inge can communicate by mobile phone. This notion is employed in Computer Science. For example.
There are some minor differences. we will use “Sachlage” and “Sachverhalt” for the items in our modeling. In the previous subsection.A Tractarian Universe Sachverhalte” for representations of Sachverhalte. In this light. is so unpleasant that the awkwardness of our usage is well worth the cost. This paper Object Objectoccurrence Poleoccurrence Pole Type Bawden’s paper Type Vertex Terminal Label – Modern graph rewriting Type Vertex Port – – Under this translation. where Tractarian logical spaces are static. without dividing out bisimilarity. we discussed the similarity between linear graphs and Sachlagen.edu/handle/1721. So. Linear graphs turn out to be almost the same as (models of) Sachlagen. Even if precisely these spaces are rather clearly not a good model for natural language semantics. graph rewriting adds a dynamics and studies processes that modify linear graphs. We may translate our concepts into Bawden’s as follows. has some interest beyond the historical reconstructive question whether one can make sense of the ontology of the Tractatus. They may be viewed as all having the same type that clicks with itself. Bawden also uses bisimulation as the proper relation of equality. the poles in Bawden’s framework are chosen from a fixed set. Since Bawden’s study is from 1993.6 Linear Graphs Vincent van Oostrom pointed out to me that my modeling of Sachlagen is very much similar to linear graphs as studied by Alan Bawden in his paper Implementing Distributed Systems using Linear Naming (see: http://dspace. he has clear priority for a number of ideas. mit. His explanation of why this is a good choice is very much similar to mine. meaning: when two representations of Sachlagen are bisimilar. what we are doing here can be viewed as a study of unsaturated objects under radical simplifying assumptions. This convergence also shows that we are looking at a natural idea. We will say things like when two Sachlagen are the same. Finally we note that. .7 Significance of this Paper I submit that the study of these logical spaces à la Wittgenstein. the idea of nonsaturated objects is a good one—and one that has not been studied sufficiently seriously. 1.1/7085). 1.
We demand that M is symmetric.10 Thus. be a theorem (Theorem 4. we can have different physical molecules of the same type. In contrast. p). and we assume that the ingredients of Wittgensteinian logical spaces can be taken to be just sets. A function P that assigns to each object a (possibly empty) set of poles. Two objects d and d have the same logical form if there is a typepreserving bijection between Pd and Pd . 5] and [6]. where Oσ . [4. e. Different atoms of the same type may occur in different molecules. functional and symmetrical and that (d. Secondly. i. is the set of (d.3) that the totality of Sachlagen can be represented by a set.A.. 10 This is a good point to note some disanalogies with the molecules of physics. It will then. e. where i is in Id and p is in Pd . A typing function F on D such that Fd : Pd → T.. It is important to stress that we are not aiming at a foundational presentation. A logical space is given by: • • • • • A nonempty set of objects D. more refined treatments of argumentplaces are studied in the philosophy of neutral relations. building up the Tractarian world and its metatheory in tandem. i. The matching relation tells us which types may combine. Of course. We model occurrences by assigning to each object a set of indices. but we will refrain from adding such extra features in the present implementation. Even if the precise identity of these poles is immaterial. Visser 2 Let There be Sachlagen In this section we develop our (class of) model(s) of the Tractatus universe. We demand that R is total (in the sense: ∀x ∃y Rxy). See. they still are like fixed argumentplaces or argumentroles. p) R (d . a tractarian object may occur in several Sachlagen and may occur a number of times in the same Sachlage. a Sachlage σ is given by: • • A function I from objects to (possibly empty) sets of indices.9 Each pole has a ‘type’. pole pole A binary relation R on Oσ . say ZF with urelements. We turn to the definition of a Sachlage in a given logical space. the set of pole occurrences. A type of atom may occur a number of times in the same molecule. in physics. adding a group of permutations to the set of poles to make certain places ‘the same’. A set of types T.g. a more flexible implementation would be possible.g. In contrast every Sachlage is unique. p ) implies Fd ( p) M Fd ( p ). e. A binary matching relation M between types. 9 Such . the same atom (numerically) cannot occur in different molecules and it cannot occur a number of times in the same molecule (at the same time). i .g. Every object comes equipped with a set of poles. Firstly. We simply employ the usual set theoretic metatheory.
j . We represent such an arrow as p. Moreover.q p. i ) :⇔ d = d and i Bd i . p) R (e. 1): • • If i Bd i and (d. q). Remark 2. p) R (e. If i Bd i and (d. i . p. i. for some j in Ie . One pleasant alternative way to think about R is as a set of arrows α between objectoccurrences of the form ((d. the set of occurrences of that object. Fig. p ). We turn to the definition of sameness of Sachlagen. We only allow poles to click together if their types match.q (d. j. Note that. for some j in Ie . It is useful to also have a click relation at hand that ignores the poles via which occurences of objects are connected. we have j Be j and (d. modulo bisimulation we can always replace such selfclicks by nonselfclicks. p) R (e. The relation R tells us which poles click together in the Sachlage. p (d. excluding it will complicate our notion of bisimulation collapse. in an alternative presentation. where i ∈ Id . However. We write (d. The demand that R is total embodies the idea that Sachlagen are saturated. q). i) −→ (e. j )). i) R (d . then.A Tractarian Universe The function I gives us. i . This seems somewhat unnatural. j . q. i) is the domain of α and (e. p) R (d . j ) is the codomain of α. i ) :⇔ ∃ p. This happens for example in the canonical unraveling. j. i). if we want to stress the symmetry of R. See Section 4. then. the set of pairs (d. q) (zigproperty / forward property). p) R (e. 1 The Zig and the Zag property . i) B◦ (d . such that Bd satisfies the zigand the zagproperty (see Fig. i). for each object. Here (d. i . i.1 We note that we allow a poleoccurrence to click with itself. j ). we have j Be j and (d. (e. i) —– (e. we could view a bisimulation as obj obj a relation between the sets of object occurrences Oσ and Oσ . obj The set of occurrences of objects in a Sachlage σ is Oσ . we define: • (d. q) (zagproperty / backward property). We will identify as many occurrences as is reasonable. i. j ). Thus. We identify all occurrences that we can consistently identify given that we want their environments to be similar. So B of our ◦ official definition would be represented by B with: • (d. We say that B is a bisimulation between σ and σ iff B is a function on objects that assigns to each d a relation Bd between Id and Id . It is like a hand that shakes itself.
2). (See Fig. Each Sachlage σ in its exclusive role represents a possible world. it follows that (e. j . Remark 2. for the range. or σ σ . We have: B is a bisimulation from σ to σ . the relation Bd is total. iff there is a total and surjective bisimilation B from σ to σ . q). j ∗ . if we want to exhibit the witnessing bisimulation B.d . ˘ b. Suppose i Bd i and (d. j . for each d. there must be e∗ . p) R (e∗ .A. By the zig property. Since σ is saturated. For any bisimulation B from σ to σ . It is surjective if. So. for each object d. Since. p) R (e∗ . and. We will show below that this implies that there is a maximal existing Sachlage. R is functional. or σ is a fact of σ . A bisimulation B is total if. p) R (e. iff there is a total bisimilation B from σ to σ . Let idσ be given by: idσ. or σ ≡ σ . i . we say that σ exists in σ . if σ σ. q∗ ) = (e.3 a. j . . let B be given by the stipulation that ˘ ˘ Bd is the converse relation of Bd . q∗ ). It will be pleasant to move in a flexible way between B and B◦ . q∗ ). similarly. i. q). Visser Fig. j . j . Consider Sachlagen σ and σ . Here are the basic facts about bisimulations.2 A surprising aspect of our framework is that the zig property implies the zag property (and vice versa). i. The Sachlage σ is a subSachlage of σ . We will write B : σ ≡ σ . We will write B : σ σ . The Sachlage σ is equal to σ . we find that. We have: idσ is a total and surjective bisimulation from σ to σ . q∗ . The domain of B will be the domain of B◦ . if we want to exhibit the witnessing bisimulation B.d is the identity relation on Iσ. When viewing σ as a possible world. we have j ∗ Be∗ j and (d. j ∗ . j ∗ witnesses the backward property. 2 Zig implies Zag It is easy to regain B from B◦ . Suppose we have the zig property from σ to σ . Theorem 2. We impose the reasonable constraint that the existing Sachlagen (in their inclusive role) are closed under sums. for each d. Here is the argument. such that (d. for some. Bd is surjective.
p) }. Let B be a set of bisimulations from σ to σ . is a partial preordering with ≡ as its induced ˘ Proof For example. The object d has one occurrence. There are classmany other representations consisting of sums of disjoint copies of the ones just given. say 0. we have Pd := { p}. we will introduce the bisimulation collapse and the canonical unraveling. The object d has two occurrences. Corollary 2. then B : σ σ and B σ. d.r. In both spaces we just have a single object d.4 The relation equivalence relation. (d. B by: (B. The second representation looks as follows.A Tractarian Universe c. Here Iesse. We define a Sachverhalt as an atom of .ens = ∅ and ⊥ Rnil is the empty relation. (d. ˘ :σ Conversely.d ⊥ relation. if B : σ ≡ σ . Pens := ∅. and. the relation M is empty. blocked from participating is a Sachlage. Example 2. The space S1 has as Sachlagen ⊥ and a second Sachlage solo.w. 0. solo has just two connected representations. 1. 0. See Theorem 3. a }. p). We just have a single object ens. In both spaces. The theorem gives us immediately the following corollary. (d. We will show that there is also a maximal ⊥ Sachlage . of . consequently. 0. 0. We have: B. Clearly. Its ≡equivalence class ⊥ of ⊥ contains precisely one element. The object ens has no poles. Our space has two Sachlagen (modulo ≡). This means that σ is a Sachverhalt σ . The space has no types. p). Let B be a bisimulation from σ to σ and let B be a bisimulation from σ to σ .t. We define the empty Sachlage. p). First we have the empty Sachlage nil or ⊥ with Inil. Modulo ⊥ isomorphism. we have σ ≡ σ or σ = ⊥ ⊥. Both spaces have one type a. .ens = {0} and Resse is the empty relation.5 Consider the following logical space BASIC. The proof of the theorem is entirely routine. 1. if B0 : σ σ and B1 : σ σ . Modulo isomorphism. for all d and R⊥ is the empty ⊥. We def ine B. We take MS0 := ∅ and MS1 := { a. for any σ We end this section by presenting some examples. p) }. then (B0 ∪ B1 ) : σ ≡ σ . say 0 and 1. The clickrelation is defined by: Rsolo = { (d. our first representation is the bisimulation collapse and our second representation is the canonical unraveling. B is a bisimulation. In Section 4. Here esse is the only Sachverhalt of our space. B )d is the composition in the order of reading of Bd and Bd . Clearly S0 has only the empty Sachlage ⊥ The poor object d is eternally ⊥. We def ine B as given by: ( B )d is the union of the Bd for B in B . We have Rsolo = { (d. The first representation is the following. i. ⊥ by: I⊥ := ∅. p) . ⊥ is the minimal element w.o. iff. ⊥. Example 2. We f ind: B is a bisimulation. (d. Then there is the Sachlage esse or in which ens is present..11.6 Consider the following logical spaces S0 and S1 .
Iσ. The types are a and b.7 Let some first order relational signature be given and some nonempty set of objects X. but of infinitely many occurrences of two objects. . and Iσ. where. 4.z := 1. The second object e has two poles. r) R (e. if z = x or z = y. .2211 Even if the world is infinitely complex. . (e. Let M := {(a. Ie := ω. . . n. 0) r. p) R (e. The relation R is generated by the following stipulations: (d. . n+1) . Thus. ∗) p.A. q). so that every fact consists of infinitely many states of affairs and every state of affairs is composed of infinitely many objects. . where r is binary. . . (e. .s = 0. 1. .2211 Auch wenn die Welt unendlich complex ist. We set Fd p := a. so the infinity of occurrences is ‘real’.q (e.0 := b.q . the Sachverhalt rxy. (b. so daß jede Tatsache aus unendlich viele Sachverhalten besteht und jeder Sachverhalt aus unendlich vielen Gegenständen zusammengesetzt ist. . b). The first object d has one pole. Fer := a. (e. n + 1. Wittgenstein writes: 4.. We take D the union of and X. for z ∈ X. if s is a relation symbol not equal to r. to wit q and r. a)}. .i := a and Fx. Consider a logical space with two objects d and e. We now consider the Sachlage that is given as follows.z = 0 if z = x and z = y. . and Iσ. r) R (e. a)}. We take M:= {(a. In this example. there would still have to be objects and states of affairs.q (e. 1) r. Example 2. • • Id := {∗}. (b. We leave it as an exercise to the reader to construct an infinite Sachlage in a logical space with only one object. our our Sachlage looks like this: (d.8 In 4. 0. 0.r := 1. q). We assume that the relation symbols in and the objects in X are disjoint. q). auch dann müßte es Gegenstände und Sachverhalte geben. we show that one may imagine Sachverhalte that consist not of infinitely many objects. n − 1}. Visser Example 2. In our example we have an f inite logical space with an inf inite Sachverhalt.g.q (e. we take Pr := n := {0. Moreover Iσ. For a relation symbol r of arity n. In this example the Sachverhalte correspond precisely with what we would take intuitively as Sachverhalte over the signature. .2211. n) r. Let Fr. to wit p. E. b). . It is easily seen that this example indeed gives us a Sachlage (even a Sachverhalt) and that the maximal bisimulation of this Sachlage and itself is the identity. we take Px := 1. is modeled as σ . Fe q := b. ∗. For an object x ∈ X.
but for the fact that we consider it as a bisimulation between other Sachlagen. i.A Tractarian Universe 3 The Ordering of the Sachlagen In this section we study the ordering . σ τ iff. We call the Eσ equivalence class of (d. σ is divided in (possibly zero) nonempty connected parts. i. p) Rσ X (d . The classes bisimilar to Y are bisimilar copies of the same Sachverhalt. We write σ [d. i) Rσ (e. The results of this section often have alternative proofs using the normal form theorems of Section 4. • • A set of objectoccurrences X is closed in σ . p) Rσ (d . The domain of B◦ is closed in σ and the range of B◦ is closed in τ . Here is a first lemma. we will first prove some basic facts. We define Bσ . an atomistic. since the identical embedding of the sets of indices for an object forms a total bisimulation.d .τ is a bisimulation between σ and τ . the Eσ equivalence classes. . Lemma 3. obj We have τ σ iff τ = σ Oτ . i) in Oσ : [d. p ) :⇔ (d. i) ∈ X}. whenever (d.τ has the same graph as Bd . i) ∈ X and (d. It is easily seen that σ [d.d := {i ∈ Iσ. Clearly. (d. Moreover. i]σ .1 Basic Facts To develop some feeling for what a Sachlage looks like. τ σ implies that τ σ . that it is.τ . • • • • Clearly. – – Iσ X. i. Eσ is the transitive. Suppose σ σ and τ τ . We show a. i . (d. reflexive closure of Rσ . Eσ is an equivalence obj relation.d j obj obj iff (d. j ) ∈ X. Suppose B is a bisimulation between σ and τ (not necessarily total or σ and τ τ . We define σ X or σ X as follows.o. we will simply say that X is closed.d  (d. i]σ . complete distributive lattice.d and i ∈ Iσ X. where i ∈ Iσ X. A Sachlage σ is connected if it has at most one Eσ equivalence class.τ is B. Iσ. then (e. if. We write Bσ. We will see that each Eσ equivalence class Y corresponds to a Sachverhalt.d ⊆ Iτ.τ for the bisimulation surjective). In d other words Bσ. σ X is again a Sachlage. j ). Suppose B is a bisimulation between σ and τ (not necessarily total or surjective). It is easy to see that Bσ . i . Clearly. Bσ. modulo ≡. for every d. p ). i] is a connected Sachlage. j ) ∈ Oτ and i Bd j.τ by i Bσ . i] for σ [d. i) ∈ Oσ . Suppose σ between σ and τ such that.d and Rσ ⊆ Rτ . To formulate these we give some basic definitions. If the Sachlage is clear from the context.1 Suppose B is a bisimulation between σ and τ . Suppose X is closed in σ . 3. Some of these Eσ equivalence classes may be bisimilar to each other. for all d.
1 We have: i. i] ≡ τ [d. j. It follows that. i) and j = (σ. We define ı Bd j iff (ı ∈ Iσ. Suppose B: σ σ and σ is nonempty. whenever σ ⊥ σ . i) is in the domain of B◦ . σ ≡ σ and σ τ. Conversely. The following theorem shows that we can decompose in and ≡ in two ways.e and (d. or ı = (σ. Suppose B witnesses σ τ .3 Each Sachlage σ is the supremum of the Sachverhalte below it. j ). σ Proof Ad (i) Suppose B witnesses σ τ .A. i]. Then. (d. Theorem 3. Proof Suppose. q)). For the converse it is sufficient to show that a nonempty connected sachlage σ is an atom.i]. τ if f. we have: Bσ. i) in σ . Visser ii. for each object occurrence (d. for all Sachverhalte σ0 σ . σ [d. σ Ad (ii) We define τ by: • • Iτ . . i] σ . i) B◦ (d. p) Rτ (e. j] : σ [d.i]. then σ τ . σ τ. Suppose (d. Then.τ [d. and.d ) is empty.2 Sachverhalte We remind the reader that a Sachverhalt σ is defined as an atom of σ is not ⊥ and. Since the range of obj B◦ must be closed in σ . so τ ≡ τ . ı. σ ii. j]. i) in σ . p) Rσ (e. σ τ and τ ≡ τ . hence. ⊥ .d := Iσ. q) iff (ı ∈ Iσ.τ : σ [d. if σ ≡ σ τ . Hence. Suppose (d. σ [d. In other words: is atomistic. σ [d. Hence B : σ ≡ σ. i] is nonempty and connected. Clearly σ τ . say this is witnessed by B(d. B◦ [d. B◦ [d. Moreover.τ Y : σ ≡ τ Y and τ Y τ . j ). Then. Theorem 3. i] τ . i] σ iii.i) . q)). we have σ0 τ . Clearly B is a total and surjective bisimulation from τ to τ . or (ı = (σ. ı. σ ≡ σ [d. p) Rτ (e. j. that is: Theorem 3. 3. Since σ is not the empty Sachlage. Proof Suppose σ is an atom. so B is surjective. for some σ . We note that Iσ.d ∪ ({σ } × Iτ. there is an object occurrence (d. for some τ . i. We leave the easy proof to the reader.d and ı Bd j).d and j ∈ Iσ. Let Y be the range of B◦ .d ). τ if f. then σ ≡ ⊥ or σ ≡ σ . j ) and (d. j. σ τ . Clearly. it must be equal to Oσ .d ∩ ({σ } × Iτ.2 We have: σ is an atom if and only if σ is bisimilar to a nonempty connected sachlage.
(σ . We define (σ.τ will be a total bisimulation between σ and τ .e j0 and j0 Be (σ . i . i) be any object occurrence in σ and suppose i Bd (σ . i ). σ σ + . Let B0 be a bisimulation from σ to σ [d. for each σ ∈ S . It is easy to see that B witnesses that σ σ . We define B between σ and σ . We are mostly interested in properties of modulo bisimulation.τ is a partial bisimulation between σ and τ .d := {(σ.3. We set: i Bd (σ. We define S =: σ + as follows.d }. for every set of Sachlagen S . (σ.A Tractarian Universe Each Bσ. σ σ . p ).i) 3. p) Rσ (d . then. Let S be a set of Sachverhalte. It is easily seen that B is indeed a total bisimulation. Let (d. i. p) Rσ + (d . We leave the easy proof to the reader.3 The SumOperation We show that every set of Sachlagen has a supremum. Then. for some σ ∈ S . p ) :⇔ σ = σ and (d. if. by j Be j iff j B0. Proof Consider any Sachverhalt σ . i). (d. Note that we do not have enough information yet to construct the maximal Sachlage as the sum of all Sachlagen: we do not yet know that we there is a set of representatives that contains a representative for each Sachlage (as an object modulo bisimulation). Let S be a set of Sachlagen. Suppose Bσ : σ τ . We define a bisimulation B from σ to σ + as follows.d j. It is d easily seen that B+ is a total bisimulation from σ + to τ . Theorem 3. if σ S . Let i be in Id . Theorem 3. We note that is defined as an operation on sets of Sachlagen. i)  σ ∈ S and i ∈ Iσ. So. Suppose B: σ S . We say that a Sachlage σ is completely join prime.i] . for some j0 ∈ Iσ [d. Hence σ + τ .4 The operation (modulo ≡). where ≡ is not divided out.5 Suppose S ⊆ S . is the supremum operation for our ordering Proof Consider any σ in S . however the following property is dependent on the precise implementation. (d. i). Hence σ τ . j ). Clearly the union of the (d. i ).i) Bσ. i].6 Every Sachverhalt is completely join prime. i] is a connected component of σ that is bisimilar to σ . . Remember that σ [d. We have: Theorem 3. • • Iσ + . S S. We will show that such a set exists in Theorem 4. i) B+ j :⇔ i Bσ.
If S ≡ S . the domain of B is the domain of B ◦ . v. for i = 0. We could also say: σ is the set of connected σ0 such that σ0 σ . ˘ Suppose Ci : τ σi .4 The ConjunctionOperation We can define an operation on Sachlagen that produces an infimum in several ways. so by (iii). we use (i) and (ii). and hence. for some σ0 ∈ σ . there is a σ ∈ S . We take σ τ := σ X. Note that B need not be total or surjective. for all σ ∈ S . iv. Similarly. • • σ is the set of σ [d. we find. For some discussion of the options. we prove that if σ0 ∈ σ . iii. σ0 σ iff σ0 σ . it might even be empty. suppose σ ≡ σ and σ0 ∈ σ . since σ0 σ ≡ σ .7 We have: i. each element σ0 of σ is below σ . 1. for some σ1 in σ . there is a σ ∈ S .10. Ergo. Let B be the maximal bisimulation from σ0 to σ1 . since σ0 is completely join prime. such that σ0 ≡ σ1 . such that σ0 ≡ σ1 . For (iii). σ ≡ σ if f σ ≡ σ . note that. then S ≡ S . for all Sachverhalte σ0 . such that σ ≡ σ . Moreover. then. such that σ ≡σ . 3. σ0 σ if f σ0 σ . i] such that (d. Let S and S be sets of Sachlagen. Then σ0 σ . Note that (σ0 σ1 ) σ0 . The . σ ≡ σ if f. Proof (i) and (ii) are easy. for all Sachverhalte σ0 . Similarly. modulo ≡. Let X be the domain of B◦ . We have: S ≡ S iff. C1 is a bisimulation ˘ from σ0 to σ1 . σ0 ≡ σ0 . that σ0 ≡ σ1 . σ σ. We take one such way as our official choice. From left to right is trivial. For any Sachverhalt σ0 with σ0 σ . Theorem 3. Visser We give two definitions. Suppose. ii. For the lefttoright direction. since C1 is total. see Remark 3. We prove (iv). σ ≡ σ σ . the inf imum of σ0 and σ1 . and for all σ ∈ S . Then. For the righttoleft direction.τ : (σ τ) τ. The domain C of is the domain of C0 . Hence σ ≡ σ . Theorem 3. (σ0 σ1 ) σ0 . there is a σ0 ∈ σ . We prove (v). Consider the maximal bisimulation B between σ and τ . there is a σ1 ∈ σ . We define the conjunction σ τ of Sachlagen σ and τ as follows. i) is an object occurrence in σ . σ ≡ σ. Proof We will use our convention that for any bisimulation B .8 σ0 σ1 is.A. A Sachlage functions in many respects as a set of Sachverhalte. Bσ τ. We note that C := C0 .
Consider any bisimulation C between σ and τ . We have: σ0 σ S ⇔ σ0 σ and σ0 S ⇔ σ0 σ and. . q) iff (d. j ). Moreover. T. This suggests the following τ := (B). since B is maximal.7(v) Remark 3. F) are just sets. q) iff (d. then Clearly C◦ ⊆ C ◦ implies definition of .d := {(i. Note that it follows that: (d. q). σ  σ is a Sachverhalt and σ τ }. It follows that any Sachlage can be modeled as a set. We may conclude that the range of C0 is included in the domain of B. where B is the maximal bisimuand τ . . {σ σ  σ ∈ S }. Proof Let σ0 be a Sachverhalt. (C) σ and (C) τ .10 There are several alternative choices for our operation Perhaps the two most obvious are: • • σ σ τ := τ := {σ {σ σ σ τ }. (d. and A third alternative is in in the spirit of our chosen definition of connects this to a beautiful observation: bisimulations can themselves be considered as Sachlagen. (i. for some σ ∈ S . j )  i Cd j}. whenever ˘ (C0 . we really do our summation over a set. (i. σ0 ⇔ σ0 {σ σ  σ ∈ S} The desired conclusion follows from Theorem 3.τ. We define a Sachlage (C) by: • • I (C). j . i. p) Rσ (e. σ0 ⇔ for some σ ∈ S .5 The Sachlagen form a Set We assumed that all the ingredients of what makes a logical space (D. (i . j ). C1 ) ≡ ν. Ergo: C0. i . We define σ lation between the Sachlagen σ 3. p) R (C) (e. p) Rτ (e. Note that in both cases. We can now easily show that C0 : ν σ and C1 : ν τ . σ0 σ and σ0 (σ σ) σ σ ⇔ for some σ ∈ S . j. j ). We prove distributivity of Theorem 3. p) R (C) (e. (i .σ0 σ1 : τ (σ0 σ1 ). P.A Tractarian Universe ˘ domain of C0 is the range of C0 . M. j ).9 We have σ over S≡ . (C) (C ). q).
making the Sachlage modulo bisimulation into a Boole Algebra. It is not the Boole Algebra that Wittgenstein defined over logical space. 4. Hence the Sachlagen form a set modulo bisimulation (Theorem 4. We clearly can tell a story. in a sense.g. In Section 4. (See e. [3]). Let B be the set of all bisimulations between σ and itself. Sachlagen and bisimulation on Sachlagen are more like graded transition systems and counting bisimulation. Visser However the totality of the Sachlagen is a proper class. our Sachlagen are more constrained than most kinds of transition systems the normal forms are. better.11 In any logical space. for canonical unraveling in modal logic one needs to choose a cardinal. τ } := ↑σ ∩ ↑τ . This is unnecessary when one studies graded modalities and employs counting bisimulation. The last kind. to make sense of this.3).11 Our sum corresponds to Wittgenstein’s conjunction: ↑ {σ. As a consequence of this result.12 We can also define subtraction of Sachlagen.g. the sum of arbitrarily many copies of a Sachlage is again a Sachlage. 4 Normal Forms for Sachlagen In this section. canonical bisimulation collapse. A Sachlage σ in the inclusive role reappears as a set of Sachlagen in their exclusive role as ↑σ := {τ  σ τ }. using the normal forms of the next section. R) as follows. consists of the bisimulation collapses of the canonical unravelings.1 Bisimulation Collapse We can find a minimal representation of a Sachlage σ := (I. . e. since we have put no constraint on the choice of the occurrences. we would even obtain a proper class if we could pick an element of every isomorphism type of the Sachlagen..g. The first two kinds of normal forms are the usual normal forms familiar from e. we have a maximal Sachlage .g. Remark 3. we provide three kinds of normal form for Sachlagen. E. we will produce a unique normal form modulo bisimulation for each Sachlage. 11 I am ignoring the fact that we do not have a set here. The Sachlagen modulo bisimulation ordered by form a atomistic distributive complete lattice.A. As we will see these normal forms form a set. we find: Theorem 3. Since. This set is not empty. The last two kinds make a bisimulationfree approach to Sachlagen possible. modal logic: bisimulation collapse and canonical unraveling. We note that this Boole algebra is the Boole algebra of Sachlagen in their inclusive role. Wittgenstein’s algebra is defined on sets of Sachlagen in their exclusive role. Since.
d1 . Moreover we demand: Fdk ( pk1 ) M Fdk+1 ( p(k+1)0) . d (d. Let B† := B . E. . We note that the result is again a path. By Theorem 2. . for some d. We call these sequences paths. . We write [i] B† for d the B† equivalence class of i in Id . . . d2 . for some object d. p . e) (e) (e. q). even if both (d. the bisimulation collapse of such a canonical unraveling is a normal form that is both bisimulationminimal and uniquely determined for each Sachlage.). f˘ is the result of reading f backwards. p. then σ0 and σ1 are isomorphic. e) (e) and (e) (e. . a path of length 0 has the form (d). p. p. we † † find that. . . These normal forms are unique simpliciter. d) are. j0 ∈[ j] B† (d. . for k > 0. p10 . q) :⇔ ∃i0 ∈[i] B† . We consider (nonempty) sequences f of the form (d0 . both f and g start from d. . . .) and g = (d. We define: • • • f # g iff. . i. .) = (. Since the objects form a set and the poles of an object form a set.A Tractarian Universe since idσ is in it. (d. The paths with form a partial monoid with as units the paths (d). . U contains the path (d). p. As soon as we have these. .) and p = p ). [ j] B† . p) R† (e. So. p. i0 . So B† is the maximal bisimulation of σ with itself.3. Moreover.. pk0 and pk1 are in Pdk and pk0 = pk1 .2 The Canonical Unraveling Consider a logical space. and. 4. we will develop normal forms using canonical unraveling. . j0 . dn ). In the next subsection. whenever σ0 ≡ σ1 . . p11 . Here p01 is in Pd0 . Warning: the paths do not generally form a category. . All elements of U have starting point (domain) d. p) R (e.g. If our operation is defined. A set U of paths is coherent iff. p01 . d) (d. d. d) is not defined. e e d d d It is easy to see that σ † is indeed a Sachlage and that σ † ≡ σ . and either f = (d) or g = (d) or ( f = (d. p20 . . pn0 . The length of a path is the unique n such that the length of the path qua sequence is 3n + 1. the paths form a set. ii. [i] B† . . each B† is an equivalence relation. f g is defined if f˘ # g. . . we have: f g (. We define coll(σ ) := σ † by: d • • † Id := Id /B† .
.i) . q . (d. then k is 3k0 + 1.i) as follows. f . In other words. i))) to σ with (d. e ) is in U. f. such that G((d. The case starting from (d) is similar. also the σU form a set. Theorem 4.(d. Visser iii. (d)) = (d.i) ) from svh(cs(σ. and a functional bisimulation G (or. U is closed under initial subpaths. . i) in σ . such that f (e. such that f = h f0 . we can f ind a coherent set cs(σ. we have (e. (d)) B◦ (d. for any bisimulation B from svh(cs(σ.1 Given any Sachlage σ and any object occurrence (d. (d. (d)) to (d. q ) Rσ (e . v. q) RσU (e . Then. (e. . j . q . We easily see that σU is a Sachverhalt.e := { f ∈U  cod( f ) = e}. either one of the paths (weakly) extends the other. q ). (d. 12 We note that this could also be formulated as follows: consider two different paths f and g in U. r . e ). assuming that the fairy would assume the Sachlage to be stable. it is extended in U. there are h. g = h g0 and f0 # g0 . i). By recursion on path length we simultaneously define elements of U and a mapping G0 on these elements such that G0 ( f ) ∈ Icod( f ) . q ) iff f = f (e . q. or. i). for e = cod( f ). q.(d. q . there are r and e . f ) G (e.A. (d. In other words.(d. Let k be the smallest number such that either (( f )k is defined and (g)k is undefined) or (( f )k is undefined and (g)k is defined) or (( f )k and (g)k are both defined and different). e) is in U and r is a pole of e unequal to q. Proof We define U σ. G0 ( f )). j. Consider any coherent set U. i))) to σ that connects (d. i) follows with an easy induction on path length. e) and G0 ( f ) =: j are already given. Then. We add f := f (e. Suppose f = (. Moreover. . . e . i))) to σ . for some unique j . in the sense that backtracking leads to the same places. f0 . i))) to σ . We treat the case that f = (. . i)) := U σ. We take (e. more explicitly. Gσ. For any f and g in U. That G◦ is part of any bisimulation from svh(cs(σ. q . r. q. for some k0 (assuming that we start counting with 0). e) or f = f (e. e ) to U. We note that. the totality of the coherent sets of a given space forms a set. when a path can be extended. after sharing an initial part they diverge by taking different poles. Suppose q ∈ Ie \ {q}. We put (d) in U and set G0 ((d)) := i. We associate a Sachverhalt svh(U) := σU to U in the following manner. g0 . (d. A coherent set is something like a map that a fairy living in a Sachlage would make of the Sachlage. . q. Since the coherent sets form a set. we have G◦ ⊆B◦ . It is not difficult to see that G is a functional bisimulation from svh(cs(σ. q . and that the fairy would never unnecessarily identify places. and take G0 ( f ) := j .12 iv. We define: • • IσU .
. unr(σ ) ≡ σ . h. i) is an objectoccurrence in σ . i)) is a subset of U and f B G0 ( f ). e) is in W and G0 ((d. Theorem 4. we have (d. . p.(d. since the σU form a set. q. p)Pσ (e. . However. i))). . since is essentially the subset relation. that the σU form a set. . i). e) is in U and (d. we see that all Sachverhalte are bisimilar to σU . q. Proof Let G := Gσ. there is a bit more to say about our construction. . (e. . p . . then unr(σ ) = unr(τ ). . Proof Our ordering is atomistic and every atom is bisimilar to a σU . This gives us (a). q) and h B j.i) and let G0 be the corresponding function of paths. the objects of the form unr(σ ) form a set. j. say X . q. q. . q. Since. q). The material we developed up to this point is sufficient for the purposes of our normal form theorem. Gσ. p) RσU (e. (d)) B◦ (d. c. e)) RσU (e . of course. q. . . . . The canonical unravelings unr(σ ) form a set. then σ ≡ svh(cs(σ. b. i. the subsets of X form a set. p. . This uses. e)) B◦ (e. e)) = j. .4 Let U be a coherent set starting with (d). p. e )) = j . p.(d. i] and since σ is a Sachverhalt. q . Then U = cs(σ. • unr(σ ) := {σU  U is coherent and σU σ }.2 Suppose σ is a Sachverhalt and (d. For each Sachlage σ . Clearly. p . q. Hence. Finally. (d. So (d. . . unr(σ ) ≡ σ . q. e) B j. (. p. q. and (c) is immediate from Theorem 3. Suppose (.3. (d. i)) and G◦ σ. Proof Since. (d). Suppose (d. Moreover. Suppose we have the induction hypothesis for (. q. e). we find σ ≡ svh(cs(σ. e ) is in W and G0 ((. We are now in the position to give our canonical unraveling of σ . Theorem 4. for some coherent U.A Tractarian Universe Theorem 4. (d) is in U and (d) B i. h . The only possible such h is (d.(d. e. If σ τ . q . Suppose B is a bisimula◦ tion from σU to σ with (d. we find that (e. We have the following theorem. (b) is trivial.5. we find that (d. (d. j ). d. e. (d)) B◦ (d. Then. q. (d. i). then unr(σ ) unr(τ ). We show by induction on path length that W := cs(σ. (. p ) and h B j . i))). . e)) = j. e). Hence.i) ⊆B . Let G((.i) is surjective to σ [d.3 We f ind: a. for some h. Thus. If σ ≡ τ . Since (d. We note that our earlier result that the Sachlagen modulo bisimulation form a distributive atomistic complete lattice follows immediately from Theorem 4.
. It is easily seen that h G (e. i)) and (τ. G ) (d). Let G := G Theorem 4.4 to σV and G. we also have the following theorem. (d. we find that U = cs(σV . then U = cs(svh(U). All this is trivial by the preceding results on coll and unr. By Theorem 4. G is the identity bisimulation of σV . Moreover. q . So. (d. Similarly. since σU ≡ σV . Visser for some h .4. We show that the situation is a bit better: σU ≡ σV . (d. Thus. . G is the identity bisimulation of σU . j )) the bisimulations B (not necessarily total or surjective) such that (d. 4. our normal form contains only isomorphic copies. j ) (and. does imply σU is isomorphic to σV .7 Suppose σU ≡ σV . ˘ σV .4 for a third time. .5 If U is a coherent set starting with (d). p . p . q. (. q . . e. (d) (G. So each ≡equivalence class has a unique normal form. . . So.3 The Canonical Collapse The canonical collapse is obtained by collapsing the canonical unraveling. Consider the category with as objects the pointed Sachlagen (σ. . σU is isomorphic to σV . q. Remark 4. then ccoll(σ ) = ccoll(τ ). p . i)) and as morphisms between (σ. . We may conclude that W ⊆ U. (d)) for (d) the starting point of U. (e. However.(e. does not imply σU = σV . for each Sachverhalt. Proof Suppose B witnesses that σU is bisimilar to σV . Moreover.6 The reader who has some knowledge of category theory.(d. q. we define the canonical collapse of σ by: • ccoll(σ ) := coll(unr(σ )). hence. . Suppose U starts with (d) and (d) B h. It is immediate that W = U. q . will suspect there is an adjunction here. h)) and G := GσV . . This collapse is evidently bisimulation minimal. (d. (e. Since p = q.h) is a functional subbisimulation of B. . Our normal forms still ‘contain’ a number of bisimilar copies of a Sachverhalt. Consider as second category the coherent sets with as morphisms just the identity. h = (. (d. where svh+ (U) := (svh(U). Theorem 4. . e. e ). . h)). G . e ) B j . We note that G.A. q. That is indeed the case. As an immediate corollary we have: Theorem 4. e). h must be an extension of (. e ) is in U and (. G is a functional bisimulation from σU to σU . d = e). Ergo. (d))). e. Then svh+ is the left adjoint of cs. Applying ˘ ˘ we find that V = cs(U. i) B◦ (e.h) . we have: if σ ≡ τ . (e)). we find that G. ˘ Suppose that V starts with (e). Applying Theorem 4. We may conclude that σU and σV are isomorphic. Then.
Hence. (σU . pole pole A binary relation R on Oγ . pole Note that with every object d there corresponds a configuration [[d]] given by (i) Id := ∅. [(σU . 5. g)) is an objectoccurrence in unr(σ ). Then. (d. We demand that R is functional and symmetrical and that (d. if d = d. i. (σV . f )) B◦ (d. and. (σV . f ) Cd (σV . and Id := {0}. • • A function I from objects to (possibly empty) sets of indices. f )] Bd . and (ii) R is the empty. g). Fγ (d. f )) is also an objectoccurrence in unr(τ ). Proof Suppose σ τ . f )] Bd . g) ∈ [(σU . (σV . So a configuration is a Sachlage minus totality. i .d . (σU . σU is an isomorphic copy of σU and σV is an isomorphic copy of σV . (σU . σU σ . i. f )) is an objectoccurrence in unr(σ ). by Lemma 3. as desired. It follows. p). p) := Fd ( p). if d = d.A Tractarian Universe Theorem 4. f )] Bd . (d. g)) are bisimilar according to C◦ restricted to unr(σ ) and since B is maximal on unr(σ ). and. Consider ı in Iccoll(σ ). p) ∈ dom(Rγ )}. f )) and (d. p ) implies Fd ( p) M Fd ( p ). . (σV . f )]Cd = [(σU . improvements and extensions of the framework.8 Suppose σ τ . σV σ . f )]Cd = [(σU . 5 Perspectives In this section. i. It follows that the occurrence [(σU . where (d. Clearly. But. where i is in Id and p is in Pd . where Oγ . that the Sachverhalte σU := (unr(τ ))[(d. A conf iguration γ is defined as follows. ccoll(σ ) ccoll(τ ). The occurrence ı is of the form [(σU . f ))] and σV := (unr(τ ))[(d. σU is bisimilar to σV . (σU . So it is sufficient to show that [(σU . p)∈Oγ  (d. We define: • • Pγ := {(d. g))] are bisimilar. We may conclude that (σV . hence. It follows that (d.1 Configurations What kind of structures do we get when we drop the demand of saturatedness? Let’s call these structures conf igurations. Let B be the maximal bisimulation on unr(σ ) and let C be the maximal bisimulation on unr(τ ). p) R (d .1(iii). is the set of (d. the set of pole occurrences. i. Since. f )] Bd . f )]Cd is in Iccoll(τ ). we briefly touch on possible variations. i. Suppose (σU . Thus the order structure takes a very simple form on canonical collapse normal forms. we find (d. hence. g)).d . Ergo. (σU .
If we multiply the poles of an object it is always in a coordinated way: we copy the whole set.and the zagproperty: • • If i0 Bd0 i0 and (d0 . E. i0 . there is a lingering doubt: is the framework we developed a ref lective equilibrium? There is a feeling that configurations are something like complex objects. However. for some i1 in Id1 and g1 in 0 g Gd1 . we implement the most straightforward idea of what the places are. i1 . It is clear that we can build in refinements of this treatment rather easily. i1 . This is not true anymore for configurations.2 Argument Places As we already noted. it is clear that some groups of poles will be linked in multiplication or contraction. we could imagine mapping objects to configurations. the fixed store of objects in the tractarian universe can be best conceived of as a store of possible objects. i1 . 1 g If i0 Bd0 i0 and (d0 . Thus. First it is hard to see how it coheres with ‘the world being all that . We define γ γ iff there is a total simulation from γ to γ . such that Bd satisfies the zig. Note that. this has several problems. we have i1 Bd1 i1 and (d0 . p1 ). i0 . We count γ and γ the same iff there is a total and surjective bisimulation between them. when we think about a morphism between logical spaces. E. However. g1 p1 ) (zagproperty). then. for the maximal existing Sachlage σ . We define a bisimulation as before. Also.g. We can say that an object d exists if [[d]] does..3 Space and Time How do time and space fit into a Wittgensteinean universe? One option would be to take them to be external to the Sachverhalte. p0 ) R (d1 . Perhaps we can improve the notion of bisimulation by taking B as a function on objects that assigns to each d a ternary relation Bd between Id and Gd and Id . However. but some groups are not. i0 . Associated with configurations. 5. g0 p0 ) R (d1 . the poles of a configuration γ do not quite behave as the poles of objects.. on each set of poles Pd we can have a group of permutations Gd . then. for some i1 in Id1 and g1 in 0 g −1 −1 Gd1 . We note that we can define the existence of a configuration γ as the fact that γ σ . Can we generalize the notion of object in such a way that objects and configurations behave in the same way? 5.g. p1 ).A. we have i1 Bd1 i1 and (d0 . Visser We define a simulation B from γ to γ just as we defined a bisimulation before. g0 p0 ) R (d1 . that come into existence if they occur in some fact. on Sachlagen the induced equivalence relation of coincided with the presence of a total and surjective bisimulation. on Sachlagen our new coincides with our old . i1 . 1 g I do not know whether this works. I do not think that we can treat all examples produced by Kit Fine in this way. p0 ) R (d1 . if we take a bisimulation variant of a configuration. now only asking for the forward property. However. i0 . g1 p1 ) (zigproperty).
Modeling relations. CSLI lecture notes (Vol. 241–253. & Visser. it would be good to find an appropriate enrichment. Acknowledgements I thank Jaap van der Does. G. 6. I am grateful to the anonymous referee for his/her helpful suggestions. Stanford: CSLI. K. Introduction au “Tractatus logicophilosophicus”. (2008). There’s just nothing connecting the moments. 37. (1988). The Review of Symbolic Logic. 4. Journal of Philosophical Logic. References 1. Wittgenstein. Göran Sundholm. It is plausible that John will have a temporal location argument place and that angry will carry one too too. Again. consider John is angry. then it difficult to understand how one could seriously speak of time at all. Secondly. An argument for Finsler–Aczel set theory. J.. J. this gives us the usual problems about how to model. P. A. Mind. The present framework does not seem to have means to enforce this idea.g. Nonwellfounded sets. Cambridge: Cambridge University Press. Martin Stokhof. Leo. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use. and the members of the Utrecht Tractatus Reading Group for their comments and criticisms. M. 1–33. (2000). Vincent van Oostrom. The Philosophical Review. 8. If we put spatiotemporal locations as objects inside the Sachlage. But inside the Sachverhalt we would like these places to be filled with the same location. 353–385. distribution. A companion to Wittgenstein’s Tractatus. so to say. 14). Ludwig Wittgenstein. For an acknowledgement of priority to Alan Bawden. 2. M. (1971). Archive for Mathematical Logic. 1(3). (1961). Moreover. (2002). Jesse Mulder. London: Routledge & Kegan Paul. It is hard to swallow that temporal and spatial relations would be outside the world. Tractatus logicophilosophicus. 335–354.. A. 267–298. Rieger. Black. 5. if we refuse to put constraints on how subsequent snapshots of the world cohere. Aczel. Leo. 109(434).A Tractarian Universe is the case’. (2004). D’Agostino. Finality regained: A coalgebraic study of Scottsets and multisets. 3. see Section 1. Marion. 41. (2000). and reproduction in any medium. Paris. L. Philosophies. . e. The identitity of argumentplaces. Fine. 109.6. Presses Universitaires de France. 9. Neutral relations. Leon Geerdink. provided the original author(s) and source are credited. part of the unspeakable. (2008). temporal order. 7.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.