J Philos Logic (2011) 40:397–420

DOI 10.1007/s10992-010-9156-0

Gnosis

Marcus Kracht

Received: 16 July 2008 / Accepted: 22 September 2010 / Published online: 8 October 2010
© Springer Science+Business Media B.V. 2010

Abstract The transition from form to meaning is not neatly layered: there is no
point where form ends and content sets in. Rather, there is an almost continu-
ous process that converts form into meaning. That process cannot always take
a straight line. Very often we hit barriers in our mind, due to the inability to
understand the exact content of the sentence just heard. The standard division
between formula and interpretation (or value) should therefore be given up
when talking about the process of understanding. Interestingly, when we do
this it turns out that there are ‘easy’ formulae, those we can understand without
further help, and ‘difficult’ ones, which we cannot.

Keywords Logic · Dynamic semantics · Judgement

[...] we must recall what we discovered about the Cartesian doctrine of
the mind, namely that the mind is not a distinct container of thoughts,
but a temporally ordered structure of thoughts. To say, therefore, that
one has a thought is to say that that thought is a part of a certain totality
of thoughts. When I say, [...] “I think (that p),” I say that the thought
that p occupies the “present-slot” in a temporally ordered system of
thoughts [...]; but when I conclude “therefore I exist,” I say that there

I have benefitted greatly from discussions with Albert Visser and Kees Vermeulen during
his stay in Utrecht in 1992–1993 and with Christa Hauenschild, David Pearce, Carla Umbach,
Gerd Wagner, Frank Wolter and Heinrich Wansing in the Gruppe Logik, Wissenstheorie
und Information. Later stages of this manuscript have benefitted from the thoughtful
comments from a reviewer for the JPL, and from discussions with Hans-Martin Gärtner,
András Kornai, Udo Klein and Christian Wurm.
M. Kracht (B)
Fakultät Linguistik und Literaturwissenschaft, Universität Bielefeld, Postfach 10 01 31,
33501 Bielefeld, Germany
e-mail: marcus.kracht@uni-bielefeld.de
398 M. Kracht

is a temporally ordered system of thoughts, the present-slot of which is

occupied by that thought.
Zeno Vendler: Res Cogitans

1 Introduction

This paper is the first of two papers that deal with the problem of getting to
know the meaning of an expression.1 They unite several independent issues
whose relevance has not been fully recognised to date. One issue is the
fact that building a representation is a process which goes beyond the mere
unpacking of definitions; a process that is intertwined with another process that
I call “gnosis”, translated roughly as “understanding”. Another issue is that
semantics cannot be reduced to denotational meanings. The process of arriving
at meanings plays an essential role in non-truth functional phenomena. For
example, topic and focus articulation encode part of this “process meaning”.
None of these observations are new. The idea that the same meaning can get
“packed” differently and that the packaging is related to the processing of the
content has been made over and over (cf. [17] and references therein). Process
oriented approaches to semantics of natural language are gaining ground now
(e.g. [14, 18]). Yet, it seems to me that what is still lacking is knowledge of
the process of understanding itself and what it consists in. It clearly is not a
matter of looking facts up in a mental database. That there is more to it than
that has been clearly argued for by Husserl in his V. Logische Untersuchung
[8]. Husserl, inspired by Brentano and Frege, asked about the psychological
underpinnings of understanding. If anything, the present paper can be seen as
a (first) attempt to provide a formal model of his ideas.2
Being a model based on processes it necessarily accounts for what is known
as bounded rationality; reasoning takes time, occasionally more time than a
human can or wants to invest. But there is more. The model specifically shows
that there are sentences which are clear and nevertheless we are unable to
understand them. It is not new to argue that meanings go beyond what is
represented in the mind, Putnam has argued this before (see [15]). However,
while he argued that often we do not know enough about the exact meanings
of particular words, I claim that even when we do know exactly what the words
mean we still may be unable to grasp the meaning of what is said in its entirety.
Thus our language allows us to formulate more sentences than we can actually

1 The sequel has meanwhile appeared as [13].

2I am not a scholar of either philosopher. There has been extensive work on the philosophy of
Husserl, e.g. by Fœllesdal and Dummett. Most of the discussion however evolves around the idea
of (static!) meanings or intentional objects. In contrast, here I am concerned with the psychological
aspects behind logic, a theme which has been vividly discussed in the 19th century but less so in
later times. Even more to the point, my specific interest is in the act of judgement, which I hold to
be essential for understanding the dialectics between the subjective nature of understanding and
the (arguably) intersubjective nature of meanings.
Gnosis 399

understand; the meaning of such utterances must therefore remain opaque

unless further tools come to rescue. Even to understand logical formulae may
be impossible despite the fact that nothing about their meaning is unclear.
The limits of our mind stem in part from a particular design of our reasoning
system. In part they have to do with the problem of abstractness of content:
logical notions do not have counterparts in ordinary experience except in the
most trivial way: truth is the presence of some fact, falsity (or better, non-
truth) its absence. Logical languages allow to express facts about the world
through concepts that seem to reach beyond it. Examples of such notions are
consequence and implication. They are as I see it purely conceptual, and a
proper semantics must treat them as that. If that is right, then the meaning
of these entities have more to do with our conceptualisations than with the
facts themselves. There is no physical correlate to logical operators (except,
perhaps, for conjunction). This is reflected in the standard notion of a model.
The problem that we face is to show how we can actually introduce these
abstract concepts that the complex propositions point at. Part of the story can
be told using logical theory: we lift ourselves up using the dialectics between
proof and formula. Once we know what a proof is we can successfully establish
higher and higher logical concepts. Moreover, as we shall see, the fact that
complex propositions live at a higher level of abstraction introduces a tension
in us; for to be able to really apply knowledge about objects of higher type
we must know how to descend to the ground level. This tension is not easily
resolved. Some propositions resist our attempts to explicate them while others
are readily decoded. Moreover, as we shall see below, it is a matter of how they
are presented symbolically that determines how well we can ‘ground’ them.
Semantics has long sought to break free from the regime of pure represen-
tationalism. Attempts in that direction have been made in update semantics
[4, 19] and dynamic semantics [6]. The abovementioned [14] and [18] also
belong in this category. Also, [20] has investigated the structure of proofs as
texts. All these approaches are in my view not radical enough. I think there is
plenty of evidence pointing to the idea that meaning is better analysed in terms
of a process than a (fixed) representation.
The present paper is a logician’s attempt to provide such a psychological
basis for the process of understanding.3 It is argued that from a theoretical
point of view the human symbolic processor is limited, and that this limitation
does not so much lead to errors but to difficulties in understanding the
content of certain utterances even if their content is in principle absolutely
clear. The central notion is the act of judgement. Humans judge sentences or

3 There is a research agenda called Grounded Cognition which declares that understanding works
via enaction, thus relies heavily on the bodily experience of the world, see [1] and references
therein. While I agree with much of what Barsalou says, there is still a need to analyse conceptually
what the reduction consists in and how the reduction from higher order notions to lower levels may
proceed. For it seems to be clear that language uses a number of concepts for which grounding is
next to impossible. Thus, if grounding was all there was to meaning then we could simply not
understand these concepts.
400 M. Kracht

representations as true, or false, or nonsensical, etc. This act of judgement is

immediate and seemingly unconditional. The possibility of what is effectively
a conditional judgement is possible only because of the ability to represent the
content of implication symbolically.4

2 Enacting Meanings

The basic idea is that understanding the meaning of something means to go

through a series of judgement steps. Each of these steps consists in the mind
taking some proposition and evaluating it. This evaluation is immediate and
its success requires that the proposition be given in a way that the mind can
respond directly. By “directly” I actually mean that we know it “verbatim”,
which is to say that it has the same form as something known to us. This means
that meanings are themselves objects (say, formulae), and identity in meaning
is in the first instance simply identity in form. Thus, the assumption here is that
the mind does not use propositions but syntactic objects themselves to encode
facts (but see below in Section 11 for a qualification).
Therefore, the fact that I know directly that all natural numbers are the
sums of four squares means that I have somewhere in my mind the following
sentence stored as a fact.
Every natural number is the sum of four squares. (1)
Thus, suppose you ask me
Is every natural number the sum of four squares? (2)
Then I shall surely answer “Yes.” To understand how I get to give this answer
we need to assume some minimal ability to manipulate surface sentences so as
to be able to connect (1) successfully with (2).
On the other hand, if I do not have the sentence (1) stored as a fact I can try
to dissect your question into its parts to see if I can make any progress. In the
present case this may be either difficult or impossible, depending on my math
background. For it is far beyond the capabilities of most people to prove this
statement from the mere definitions. But maybe I know some other fact that
will make it possible to answer the question. Thus, in some way the analysis will
yield some progress and thus allow me to judge true what I did not know off-
hand or even thought I didn’t know. It is important to distinguish here between
explicit knowledge and implicit knowledge (see [7]). I know implicitly anything
that my mind can possibly reach; I know explicitly only what I know verbatim,

4 The idea to this paper reaches back as far as 1988. At that time I was trying to provide a semantics

for theme and rheme, which I learned about from [21]. Unfortunately, at that time few people in
formal linguistics were really interested in the matter. That has meanwhile changed. Although
topic and focus are now a big issue in linguistics I find the approaches too conservative in that they
are stuck with standard truth conditional semantics in one or the other form. This paper has been
the result of rethinking my earlier attempts in [11] and [12], which remained unpublished.
Gnosis 401

that is, what I know without further analysis of the sentence. The judgement is
therefore of the formula and not of the content that it expresses, though it is
intended to be a judgement of the content.5
In order to see how this is possible we need to dissect the mechanism of
judgement. Judgement has two phases. The first is the preparation phase, the
second the actual judgement phase. We prepare the judgement by presenting
the mind with the necessary pieces: the mode and the formula. When in the
first phase we bring a formula “ϕ” to our immediate attention I say that we
apprehend “ϕ”.6 When we apprehend a formula it is always in a certain mode.
Modes differ; there is a mode of truth, of falsity, of possibility, necessity, and
so on.7 Brentano [2, p. 15] even says that pastness or futurity is a mode: when
we say “Caesar has crossed the Rubicon.” the present tense does not signal
existence. Caesar’s crossing the Rubicon does not exist. I leave that matter
aside here for the purpose of this paper. What is essential is that there are
many ways of apprehending a formula. Having fixed a mode and a formula
there is an immediate response, called judgement. The judgement says whether
that proposition has the given mode. If the mode is truth, denoted by “”,
and the proposition is “ϕ” then we write “ ϕ” to say that the proposition is
judged to be true. “ ϕ” states that “ϕ” is not judged to be true. It is important
to stress that this is not the same as saying it is judged false. There is a clear
difference between an act of positive judgement, that is, saying that the formula
is false, and the rejection of the opposite judgement, which is tantamount to not
saying that it is true. The latter leaves open whether the proposition might be
judged false or undecided, merely conceivable, or something else. Specifically,
it is consistent with coming (after some reflection perhaps) to the conclusion
that the formula is indeed true. Furthermore, as we are more often than not

5 Note that I avoid the term “proposition” and use “formula” instead. For the purpose of the
gnostic process the notion of a proposition is actually not helpful. All knowledge is stored
syntactically and is recalled as such. How the meaning as an intentional object comes in is beyond
the present analysis. How for example can I assure myself of the correctness of a reasoning step in
virtue of the meaning of the syntactic primitives is not a question that can be meaningfully asked
in this context. Meanings are conspicuously absent.
6 For formulae and symbols, the quotes “· · · ” are used to denote the syntactic object as opposed to

the meaning that it has. They may be superfluous when the context makes matters clear anyway
but I like to keep them as a sort of reminder. Notice for example the difference between saying “on
condition that ϕ” and saying “on condition that “ϕ””. The former says that whatever ϕ denotes is
the condition, the latter is short for “on condition that “ϕ” is judged true” (again quotation marks
are needed).
7 One may not like the idea of a separate mode of falsity. I think of two objections. The first is that

we need to control for interdependence: if I judge “ϕ” true I should not also judge it false. The
other objection is that if I apprehend “2 + 2 = 5” for truth I will right away see that it is false even
though I have not used that mode. To answer the first objection I note that the interdependence
is rather widespread. If “ϕ” is true it should also be possible, for example. Whether or not it is
necessary to control for consistency is not clear to me. To answer the second objection I say that
the judgement of falsity occurs after we have apprehended the formula for truth and rejected
the judgement. What we have is a substitute judgement, like answering the question “Is John at
home?” by “He might be.” rather than “Yes.” for lack of better knowledge.
402 M. Kracht

uninformed whether a particular fact obtains or not, we can in most cases judge
a given proposition neither true nor false.8
Judgement is preceded by apprehension. In case a proposition is not
known verbatim we therefore reject the judgement when apprehending it.
For example, suppose we know verbatim “ϕ” and “χ” but not “ϕ ∧ χ”. When
apprehending “ϕ ∧ χ” we therefore reject the judgement. This is because the
apprehended formula is not known to us verbatim. (There is a clear analogy
here with logical programming. The database contains far less than what we
can actually prove from it.) To get consent to “ϕ ∧ χ” we therefore must
proceed through a series of steps so that at the end the expected judgement
can be made. These steps are to apprehend for truth first “ϕ” and then “χ”.
If both formulae are judged true the entire formula is judged true as well. The
judgement of “ϕ ∧ χ” is therefore not direct as in the above case. It is mediated.
Every time we see an unknown proposition we are in need of mediation in
order to reach a judgement. In this paper I show how this mediation may
proceed. As this paper is not a psychological one I cannot and do not claim that
this is exactly the way humans reason. I am mainly using plausibility arguments
and introspection in the hope that other people find this interesting enough to
pursue experimentally.
Crucially, the act of judgement is performed in time. Judgements are made
one at a time. Once made, they become history. Judgements may depend
on other judgements. In order to be able to effectively make use of past
judgements they need to be recorded. With this record we can access our past
judgements. Notice however that “ ϕ” denotes the act of judgement. It is
impossible to store the judgement in the form “ ϕ”. Rather, the only thing
that is stored is a formula. In this case it is simply “ϕ” itself.9 If the judgement
was falsity, matters would be different. “ ϕ” denotes the judgement of “ϕ” as
being false. To store this we would have to enter “¬ϕ”. I shall return to this
matter in Section 9.

3 Conditional Judgement

That we judge a formula true once does not mean that we always judge it
true. On the one hand we may change our mind about it; on the other hand
our judgement may depend on certain conditions. We consistently judge true
“John is a man.” if we know that John is a bachelor; however, if we do not
know that he is a bachelor we might or might not judge “John is man.”
true. In order to explain this I shall say that we have certain conditional

8 Note that “ ϕ” also is an act, that of rejecting the judgement. It means having apprehended the
formula in the mode “truth” and not found “truth” to apply to the formula. This is not to say that
ϕ has beend found untrue. There is no piece of notation for the nonoccurrence of the judgement. I
assume for simplicity that apprehension is followed by one of the two acts: affirmation or rejection.
9 It could also be “T(ϕ)”. The material form of the object is irrelevant; we need to fix one here for

the purpose of discussion.

Gnosis 403

judgement dispositions. If “ ϕ” is a judgement then the disposition to issue

this judgement on condition that “χ1 ”, · · · , “χn ” is written
χ1 , · · · , χn  ϕ (3)
Notice that this notation defines a behavioural disposition; it is a property that
a person A does or does not have. And it is the property that if A judges
all of “χ1 ”, · · · , “χn ”, true then A also judges “ϕ” true. In this formulation
the use of the indicative hides a difficulty. On the one hand we would like
to take this literally and subject A to a series of judgements. However, after
each judgement A’s behavior may potentially change. Furthermore, we need
to be clear about how much time may intervene between these question etc.
So, ideally we would like to read this as: if at this moment A would judge
“χ1 ” through “χn ” true, A will also judge “ϕ” true. It is the latter form in
which I define the disposition; it is the first by means of which we can test
whether A possesses it. This is why in the form of (3) the judgement sign only
attaches to the last formula. The judgement disposition is a disposition to a
single act provided certain conditions are met. The conditions are given as “χ1 ”
through “χn ”, in linear order, though presenting them as a set would perhaps
be more appropriate. Note that the disposition to unconditionally consent to
the formula “ϕ” can be seen as a limiting case of a conditional judgement
disposition, namely where n = 0.
Recall now from logic the Deduction Theorem (DT). It says that given any
set  of formulae and a single formula ϕ,
  ϕ → χ if and only if ; ϕ  χ (4)
Like in constructive approaches to logic, the DT is used as a means to define
the meaning of the arrow. Let us see how this can go.
Suppose someone wants to tell us what “→” means. Rather than using truth
tables he will probably do the following. He will say: suppose you want to know
whether or not “ϕ → χ”. So you assume “ϕ”. If then “χ” is the case, so is “ϕ →
χ”, without the assumption that ϕ. What he is giving us is not a definition of
the meaning in conventional terms. This looks more like a recipe for deciding
whether or not we should judge a formula true. And he does so by reducing
the arrow to a conditional judgement. There is a price to be paid, though. In
order to see whether we accept “ϕ → χ” we have to perform a series of steps.
I have earlier spoken about the difference between explicit and implicit
knowledge. The same distinction applies to judgement dispositions. Some
judgement dispositions are explicit others are implicit. The explicit ones are
the ones stored verbatim in our mind. The implicit ones are those that can be
derived by a reasoning process. The distinction between explicit and implicit is
not about what we know; it is about how and how fast we arrive at a conclusion.
Here is an example. Suppose a person A has the judgement disposition to
unconditionally consent to “ϕ → χ”. And suppose that a person B has the
disposition to consent to “χ” on condition that “ϕ”. If asked whether they
consent to “ϕ → χ” A will respond immediately. He will confirm without
further hesitation. B however will have to make several steps. First, he will
404 M. Kracht

apprehend the formula and reach no judgement. Whence he will try the
equivalence in (4) and reduce the problem by assuming “ϕ”. He will thus
pretend that ϕ is true. Since he has the disposition to judge “χ” true on
condition “ϕ” he will actually judge “χ” true. This judgement is conditional
on “ϕ”, so he will conclude that he accepts “ϕ → χ” unconditionally.10
The equivalence of “ ϕ → χ” and “ϕ  χ” in terms of what can be
concluded logically from them has caused a great deal of concern in philosophy
(see [5]). I do not see why that should be so, and I offer plenty of reasons
(I hope convincing ones) that the difference between them is actually an
interesting one and should not be dismissed as irrelevant.

4 Assumptions

Evidently, in order for the reasoning to work properly we must be able to

distinguish two kinds of states: one where we hold “ϕ” true (say, because we
know it off hand) and one where we simply assume “ϕ”. For if we merely
assumed that John is a bachelor, say, in order to reason that he is a man, then
we only know that he is a man on condition that he is a bachelor. But this is
not to say that we think that he is a bachelor.
Evidently, as reasoners we must be able to represent the difference, and
moreover be able to keep track of our assumptions and judgements. For once
we have reached the judgement that John is a man we would like to keep that
in mind for the sake of being able to conclude that John is a man if he is a
bachelor. The fact that the latter is a condition can only be recalled at that
moment if we know that we have made that assumption.
It now follows that the process of unpacking complex formulae involves
several things. One is that we need to be able to undo the formula into
parts and reason with them independently. Each of the parts may either
be apprehended or unpacked. In each case a further analysis is performed.
Another thing involved is knowing in which way to reason, that is, in which
way to schedule the thought process. I shall leave that out of discussion, though
it is clear that in order to obtain a desired conclusion proper scheduling is
necessary.
The reasoning process needs a structured memory of past acts. This we shall
assume to be a sequence of formulae that are either marked as factual or as
assumed. Such a sequence is called a slate. If we have judged “ϕ” true we enter
it on the slate. If we have only assumed “ϕ” then we also put it onto the slate;
now however it carries a mark saying that it is an assumption. We write
ϕ (5)

10 There is a further question in the background, namely: how does B know what to do? I leave that

question unexplored. I note only that for this we need B to have some idea of how to decompose
formulae into parts or how to apply definitional schemes. But even with all this in hand there is a
further question, namely: how does B know which way to go? I suggest that this question has no
unique answer; people differ in the way they handle the actual scheduling of the reasoning steps.
Gnosis 405

If we now apprehend “χ” and judge it true then we can also put “χ” onto the
slate:
ϕ  χ (6)
Based on this slate can now derive the judgement that “ϕ → χ” is true. The
order in which we have put them is relevant. It reflects the fact that “χ” was
judged after (and so, potentially, because of ) “ϕ” has been assumed.

5 Limits of Reduction

Using the Schema (4) we can eliminate the occurrence of certain arrows;
however, not all arrows can be eliminated. Here is an example.
 ϕ → ((ϕ → χ) → χ) (7)
Using DT twice we can reach the following form.
ϕ; ϕ → χ  χ (8)
The remaining arrow is not eliminable. This is unfortunate insofar as we would
like to maintain that if we judge “ϕ” true and judge “ϕ → χ” true then we
judge “χ” true as well. This may well be the case but it is to be noted that it
requires assumptions about our reasoning system that go beyond the mere use
of the Schema (4).
However, the definition as given makes use of the judgement sign “”, and
thus cannot be used to reduce any of the “χi ” should they in turn contain
occurrences of “→”. In other words: the definition is such that it places limits
on reducibility which are not matched by syntactic restrictions. We have not
excluded leftward nesting of the arrow, but the arrows can then not be fully
eliminated via DT. Namely, we need to be able to use the equivalence of
“ϕ → χ” with a conditional acceptance of “χ” also in case that “ϕ → χ” is
merely assumed. The difference is that when we know for a particular instance
of “ϕ” and “χ” that “χ” on condition that “ϕ” then we know this in the form
of a conditional disposition. If on the other hand we know (or have assumed)
“ϕ → χ” then our direct intuition does not yield “χ” when assuming “ϕ”,
as there is no disposition to that effect. Hence, this must be postulated as a
particular rule of reasoning (the rule is called f iring below).
Now, even with the rule of firing there are still formulae that cannot be
understood (in the sense of being judged either true or false) even though they
are classically valid. One example is Peirce’s Formula.
((ϕ → χ) → ϕ) → ϕ (9)
We use (4) once and reach
(ϕ → χ) → ϕ  ϕ (10)
Now however we are stuck. There is no rule that allows us to reason with our
assumption. (We shall actually show below that this is so.) This is an interesting
406 M. Kracht

fact inasfar as it shows that there are barriers inherent in the process of simply
unpacking formulae. These barriers result from the fact that understanding
formulae is connected with being able to tell effectively whether they are true.

6 The Calculus of Judgements

I will now present a formal account of the various notions presented so far. This
will be a model of what is going on in the mind of a person. We fix a language
L of propositions or formulae, with certain syntactic rules. L contains certain
variables and constants and operators, such as “→”, “∧”. Second, fix a set M of
phematic signs (denoting modes). A judgement is an element of M × L, that is,
a pair “ϕ”, where “” is a phematic sign and “ϕ” a member of L. Notice that
a judgement applies to the formula “ϕ”. At present the only phematic signs
are “” and “”, which stand for acceptance and supposition, respectively.
(To call supposition a judgement is stretching the intuition somewhat but I can
think of no better name.)
A conditional judgement disposition has the form “  ϕ” where  is a
set of formulae and “ϕ” is a judgement. A theory is a set of conditional
judgement dispositions. Our theory of the world is thus described by a set of
conditional judgement dispositions. There is no condition on this set; it may
even be inconsistent. The state of a reasoner is a triple T, S, A such that T
is a theory, S a sequence of (marked) formulae called slate, and A is empty or
contains one judgement. Reasoning proceeds by passing from one triple to the
next. All three components of the triple are time dependent, though we shall
keep T fixed throughout, to keep matters simple. When T is empty, it will be
dropped. Slates are short term devices to keep track of one’s acts. Once we
have made a judgement we need to store it in the slate or else it will be lost.
Hence we proceed from T, S,  ϕ to T, S ϕ, ∅ and not to T, S, ∅. This
would be a rather pointless procedure, though logically entirely correct. The
fact that S now contains ϕ means that we can retrieve ϕ for further use.
For the construction of slates we have a constructor “”.11 “ϕ” indicates
that “ϕ” is an assumption. This is reminiscent of line-by-line deductive systems
where vertical lines are drawn to symbolise the area of validity of a particular
premiss. In a slate a premiss takes scope to its right but not to its left. Thus, if
the slate has the form S ϕ  S , then “ϕ” is valid throughout S . The judgement
window A is thought to be placed at the end of the slate. Thus all assumptions
are visible for A. We shall give some basic rules of transitions. The arrow 

11 The article [9] came to my attention long after I wrote this paper. In it, Kearns actually uses a

similar notation, namely “ϕ”. Strange coincidence.

Gnosis 407

is used as follows. If σ  σ  then a reasoner can pass from state σ to state σ  .

(See Definition 1.)
Œ Assumption T, S, A  T, S,  ϕ.
 -Conversion T, S,  ϕ  T, S ϕ, ∅.
Ž -Conversion T, S,  ϕ  T, S ϕ, ∅.
 -Activation T, S ϕ  S , A  T, S ϕ  S ,  ϕ.
 -Activation T, S ϕ  S , A  T, S ϕ  S ,  ϕ.
‘ Phatic Enaction T, S, A  T, S, ϕ, provided that   ϕ ∈ T and for
all δ ∈  either δ or δ occurs in S.
’ Reflection T, S ϕ,  χ  T, S,  ϕ → χ.
“ Firing If ϕ → χ ∈ S or ϕ → χ ∈ S then T, S,  ϕ  T, S,  χ.
” Forgetting T, S ϕ  S , A  T, S S , A.
Some comments are in order. The letter A in the rules is a variable; possible
values are ∅ or a judgement. The rules of Conversion are very important:
they allow to convert a judgement into a proposition so that it can be stored
in the slate. Their counterparts, the Activation rules, allow to reactivate the
judgement. Below I shall have more to say about the general notion of
conversion.
There are two rules concerning the behaviour of implication. Reflection is
a principle that relates the structure building operation “” with “→”. Firing
allows to use the implication to reason forward with a premiss. Lastly, there
is a rule that allows to drop intermediate results (“Forgetting”). Notice that it
applies only to facts, not assumptions.

Definition 1 (Derivation) A derivation is a sequence  = σi : i < n + 1 such

that each σi = Ti , Si , Ai  is a state and for every i < n, σi+1 follows from σi via
one of the above rules. We also say that  derives σn from σ0 .

Notice that reasoning proceeds from state to state, the previous states are
inaccessible. The definition says which states are derivable from which other
states. As matters stand, every state therefore depends on a prior state. Still,
there is a notion of a state that can be unconditionally derived. To establish
such a notion we need to know whether that formula can be derived given
any starting state. Since that notion defies verification the reasoner can instead
resort to a different notion, that of an axiom.

Definition 2 (Axiom) T, ∅, ∅ is an axiom. A state is derivable if it has a

derivation beginning with an axiom.

An axiom is something of a “reset” state, where we clear temporary

assumptions.
Of particular interest are the conditional judgement dispositions of a rea-
soner. They can be immediate or derived. If   ϕ ∈ T then the disposition is
immediate, otherwise derived. It matters how many steps it takes to derive the
disposition for the more steps the more time is needed to execute it.
408 M. Kracht

Definition 3 Let  = {δi : i < n}. Write  T ϕ if T, , ϕ is derivable,
where  := δi : i < n. If  = ∅ it can also be dropped.

Proposition 4 Suppose T ⊆ T  . Then if  T ϕ then also  T  ϕ.

We say that an agent consents unconditionally to ϕ if T ϕ, and that he

consents to ϕ a priori if  ∅ ϕ. The formulae that an agent unconditionally
and a priori consents to have a special status; anyone can in principle find out
that they are true.
The calculus bears resemblance to standard calculi. It is like a Gentzen
Calculus in making explicit the context of judgement. In a Gentzen calculus
each line is a valid statement on its own. It is like a Hilbert Calculus in being
based on forward reasoning or modus ponens. It is however not a Hilbert
Calculus: the distinction between a formula and a rule cannot be made in
a Hilbert Calculus. It is also not a Gentzen Calculus because it emphasises
the simplicity of the judgement act and therefore restricts the structure of
the judgement context. The distinction between fact and assumption is also
not made in a Gentzen Calculus; the context only contains the assumptions.
Also, in addition to making the assumptions explicit it also makes the rules
explicit (via the theory). I draw attention also to [9], who aims at formalising
illocutionary logic. Kearns representations are static, they lack any flexibility,
which in my opinion is the actual engine of our reasoning facility.

7 Examples and Formal Results

We shall now look at some examples in more detail. I shall show that any agent
can be brought to consent unconditionally to an intuitionistically valid formula
in →. All formulae that are not intuitionistically valid require some nonempty
T to derive them.
The lines in these examples show the state of a person at a given time. Time
is implicit (not represented) and is thought to proceed top to bottom. The
column labelled S shows the slate at the given moment; the column labelled A
shows us the judgement act at that time. We assume T to be empty, though any
other T will do, by Proposition 4. It is therefore not shown here. The rightmost
column shows the rule that has been used; it is not part of the state.

S A
ϕ Œ
ϕ ∅  (11)
ϕ  ϕ 
ϕ→ϕ ’
We begin with some remarks on derivations. If there is a derivation of S, A
then there also is a derivation of U  S, A for any U. Simply prefix the slate
of every state of the proof with U.
Gnosis 409

Lemma 5 (Leftward Monotonicity) Let S, A be derivable. Then so is

U  S, A.

By Proposition 4, derivability also is monotone in the set of judgement

dispositions. A set S of formulae is said to be closed under modus ponens (MP)
if whenever ϕ → χ ∈ S and ϕ ∈ S also χ ∈ S.

Theorem 6 The set of propositions that a person unconditionally accepts con-

tains intuitionistic logic for “→” and is closed under MP.

Proof Intuitionistic logic for “→” is characterised by two axioms and one rule:
the axioms are FD (derived in Table 1) and ϕ → (χ → ϕ) and the rule is MP
[16]. We derive the second formula as follows.

S A
∅ ϕ Œ
ϕ ∅ 
ϕ χ Œ
(12)
ϕ  χ ∅ 
ϕ  χ ϕ 
ϕ χ →ϕ ’
 ϕ → (χ → ϕ) ’

Now we show closure under MP. Suppose the reasoner unconditionally accepts
“ϕ → χ” and “ϕ”. Then he derives ∅,  ϕ → χ and from there ϕ → χ, ∅.
Denote this derivation by 0 . Now, take a derivation of ∅,  ϕ and prefix

Table 1 Deriving Frege’s Dreierschluss (FD)

S A
∅  ϕ → (χ → ψ) Œ
ϕ → (χ → ψ) ∅ 
ϕ → (χ → ψ) ϕ→χ Œ
ϕ → (χ → ψ) ϕ →χ ∅ 
ϕ → (χ → ψ) ϕ →χ ϕ Œ
ϕ → (χ → ψ) ϕ → χ  ϕ ∅ 
ϕ → (χ → ψ) ϕ → χ  ϕ ϕ 
ϕ → (χ → ψ) ϕ → χ  ϕ χ →ψ “
ϕ → (χ → ψ) ϕ → χ  ϕ  χ →ψ ∅ Ž
ϕ → (χ → ψ) ϕ → χ  ϕ  χ →ψ ϕ 
ϕ → (χ → ψ) ϕ → χ  ϕ  χ →ψ χ “
ϕ → (χ → ψ) ϕ → χ  ϕ  χ →ψ ψ “
ϕ → (χ → ψ) ϕ → χ  ϕ ψ ”
ϕ → (χ → ψ) ϕ →χ ϕ→ψ ’
ϕ → (χ → ψ)  (ϕ → χ ) → (ϕ → ψ) ’
 (ϕ → (χ → ψ)) → ((ϕ → χ ) → (ϕ → ψ)) ’
410 M. Kracht

every slate with “ϕ → χ”. This is a derivation of ϕ → χ,  ϕ. Call it 1 .

Then 0 ; 1 is a derivation of ϕ → χ,  ϕ. Now continue as follows. With
-Conversion, we get ϕ → χ  ϕ, ∅ and so with Firing ϕ → χ  ϕ,  χ.
Apply Forgetting twice and we get ∅,  χ.

We shall show that in fact only intuitionistically valid formulae are deriv-
able. Hence not all theorems of classical logic can be derived. Put

σ (S) := {ϕ : ϕ ∈ S} (13)

Notice that while S is a sequence, σ (S) is a set.

Lemma 7 Suppose that S,  ϕ is derivable. Then σ (S) Int ϕ.

Proof By induction on the length of a proof. We shall show a stronger claim,

namely that (1) for all χ in S: σ (S) Int χ and (2) if A = ϕ then also σ (S) Int
ϕ.
For the axiom, ∅, ∅, (1) and (2) evidently hold. Now assume that they hold
for σi = Si , Ai  and let σi+1 = Si+1 , Ai+1  be the result of the application of
a rule. Assumption, -Activation, -Activation, -Conversion, -Conversion,
and Forgetting are immediately clear. As T is empty, Phatic Enaction trivially
holds. We are left with Reflection and Firing.
For Reflection we have Si = Si+1 ∪ {ϕ}. (1) is evident. We need to show
(2), which is σ (Si+1 ) Int ϕ → χ. By inductive hypothesis σ (Si ) Int χ, and
this gives σ (Si+1 ); ϕ Int χ. This establishes that σ (Si+1 ) Int ϕ → χ. Finally,
Firing. Again (1) is immediate. We need to show (2). We have Si+1 = Si . We
have to show that σ (Si+1 ) Int χ. Suppose first that ϕ → χ ∈ S. Then by (1)
of the inductive hypothesis, σ (Si ) Int ϕ → χ. Suppose next that ϕ → χ ∈ S.
Then again by IH, σ (Si ) Int ϕ → χ. Now since σ (Si ) Int ϕ (by IH) we have
σ (Si ) Int χ as well. Thus, σ (Si+1 ) Int χ since Si+1 = Si . This shows (2) also in
this case.

Theorem 8  ∅ ϕ iff  Int ϕ.

Notice that this does not apply to  ϕ; that is, we cannot show that
σ (S) Int ϕ if S,  ϕ is derivable. For the latter is always derivable. Now, if
∅,  ϕ is derivable, so is also ϕ, ∅. Hence, every unconditionally a priori
accepted formula is intuitionistically valid. It follows that Peirce’s formula is
not accepted unconditionally a priori.

Theorem 9 (Deduction Theorem) S,  ϕ → χ is derivable iff S ϕ,  χ is

derivable.
Gnosis 411

Proof Assume that S,  ϕ → χ is derivable. We continue as follows:

S A
S ϕ→χ
S ϕ → χ ∅ Ž
S ϕ → χ ϕ Œ
(14)
S ϕ → χ  ϕ

∅ 
S ϕ → χ  ϕ ϕ 
S ϕ → χ  ϕ χ “
S ϕ χ ”

Conversely, using Reflection we get S,  ϕ → χ from the last line.

We get the following exchange property for free:

Corollary 10 Let S consist of assumptions only and let S be a permutation of

S . Then S, A is derivable iff S , A is derivable.

Notice also that since we may drop all premisses of the form ϕ from S, it is
enough to study only derived dispositions of the form   ϕ.
So, what then happens to Peirce’s formula? We can try to get it by using
DT. In that case, the agent assumes “(ϕ → χ) → ϕ” and sees whether he can
consent to “ϕ”. The assumption is an implication, so he may try to assume the
premiss, “ϕ → χ”. This gets him the conclusion but leaves him with showing
“ϕ → χ”. We know of course that no matter how he turns matters around he
cannot do it. But it is enlightening to try it out anyway.

8 Different Forms of Knowledge

One of the advantages of this model is that it allows to represent the difference
between (15a) and (15b).

If ϕ then χ. (15a)

Suppose ϕ. Then χ. (15b)

(15a) consists in the single judgement “ ϕ → χ”, while (15b) consists in

the sequence of judgements “ ϕ” followed by “ χ”. For notice that every
natural language sentence contains a phematic sign. In this way the utterance
of a sentence becomes a phatic act. We do not simply make a noise or denote
a proposition, we also endow it with a force. A thetic sentence endows it with
the force of an affirmative judgement. Thus, by uttering /If ϕ then χ./ we
claim the truth of its propositional content. That is why its representation is
“ ϕ → χ”. The sequence of two sentences allows to make two phatic acts.
The first of them, /Suppose ϕ./ is actually an imperative: it asks the listener
412 M. Kracht

to assume ϕ. The representation is “ ϕ”. The second says /Then χ./. It

claims that under this condition χ holds. This can be represented as follows.
ϕ→χ (16a)
ϕ
(16b)
χ
Notice that in the second example, /Then/ really is metalogical. It announces
that you will get to the conclusion on some condition, saying: if you suppose χ
then you will find that ϕ.
Additionally, it is possible to represent knowledge in different ways. The
theory of some person may contain either the disposition “ϕ  χ” or the
disposition “ ϕ → χ”. Though by the results obtained above these two
representations are identical in what can be deduced from them, there are
differences in how easy this is. Below I present two reasoners assessing both
(15b) and (15a).

T S A
T S A
ϕχ ϕ
ϕχ ϕ
ϕ  χ ϕ ∅ (17a)
ϕ  χ ϕ ∅
ϕ  χ ϕ  χ
ϕ  χ ϕ  χ
ϕχ ϕ→χ

T S A
ϕ →χ ϕ
ϕ →χ ϕ ∅
ϕ →χ ϕ ϕ→χ T S A
(17b)
ϕ →χ ϕ  ϕ → χ ∅ ϕ→χ ϕ→χ
ϕ →χ ϕ  ϕ → χ  ϕ
ϕ →χ ϕ  ϕ → χ  χ
ϕ →χ ϕ χ

In (17a) the theory contains the simple disposition “ϕ  χ”, while in (17b) it
contains the ‘coded’ version “ ϕ → χ”. While the simple disposition “ϕ  χ”
can be used directly, the disposition “ ϕ → χ” must be recalled and then used
for reasoning via Firing. However, if used on the sentence (15a) the reasoning
is much faster. This is because the target of judgement is known to the person
verbatim.
In the other cases, however, phrasing it as in (15b) makes it easier for the
reasoner to follow. He has to do less work for each individual claim. He can
asses the validity by looking at A almost step by step.
It is an interesting question to ask how someone comes to acquire “ϕ  χ”
as opposed to “ ϕ → χ” into his theory. Basically, as we observed earlier,
“ϕ  χ” corresponds to a disposition learned by experience. Unlike the latter
it does not use any linguistic sign to encode the relation between “ϕ” and “χ”.
Gnosis 413

However, “ ϕ → χ” does just that and is typically the result of explicit (i. e.
verbal) instruction or thought process.

9 Conversion

I shall now return to the rules -Conversion and -Activation. The reason that
we need these rules is the following. The slate contains only assumptions and
facts, not judgements. Thus if we have made a judgement we need to record it
as a fact. If the phematic sign is “” we simply store the formula itself. But in
other cases we need some way to encode the character of the judgement. If the
dimension is falsity, for which we use the sign “”, we use ordinary negation.
Thus the corresponding rule is this.
T, S,  ϕ  T, S ¬ϕ, ∅ (18)
In addition, we need an inverse rule of “deconversion”:
T, S ()¬ϕ  S , A  T, S ¬ϕ  S ,  ϕ (19)
Here, ()¬ϕ means that the formula ¬ϕ is either fact or assumption. We can
also write this as
If ()¬ϕ ∈ S then T, S, A  T, S,  ϕ (20)
Notice again the asymmetry. Once a judgement has been issued we can only
put it on top of the slate, but it can be recalled from anywhere inside it. An
alternative route to record the judgement “” is by judgement conversion.
Let σ  σ  abbreviate the conjunction of σ  σ  and σ   σ . Then we can
alternatively add the following pair of rules.
T, S,  ϕ  T, S,  ¬ϕ (21)
Dropping T and S we can further simplify this notationally to
 ϕ   ¬ϕ (22)
If the dimension is potentiality the rules are as follows.
 ϕ   3ϕ (23)
The general form of these rules is this. For each judgement sign “” we need a
corresponding unary operator “O ” together with the pair
ϕ   O (ϕ) (24)
The conversion rules can also be seen as rules explaining the content of the
operator “O ”. They connect a phematic sign with some expression. Likewise
we need rules that define the meaning of ordinary expressions like /man/. Once
we have an internal correlate to this expression, that is, once there is a concept
of manhood that we possess, we like to connect it to the word /man/. The way
to do this is by a similar rule of judgement conversion:
 man(x)   man (x) (25)
414 M. Kracht

This says that we judge x “man” just in case we judge x “man ”. It is standard
to call “man” the exponent (or signifier) and “man ” the meaning. However, it
is not possible to say this on the basis of the rule. The rule is ignorant about the
distinction between signifier and signified. What makes “man” an exponent
is merely the fact that it is part of English while “man ” is not. This must
be supplied elsewhere. The rule system itself allows to mix expressions and
meanings.12
Furthermore, I can link an expression of one language to an expression of
another. I can have as part of my judgement dispositions the following rule
relating the Hungarian expression /ember/ to the English expression /man/.
 ember(x)   man(x) (26)
On the other hand I could also have the following rule:
 ember(x)   man (x) (27)
In fact, I could even have both of them. All of these options seem to correspond
to reality. Speakers that have learned a foreign language like to translate the
sentences of that language into their own before they can access the meaning.
It is therefore suggested here that what they possess is not a direct link from
words of the foreign language to the concept but rather to their counterpart
in their own language, though they might acquire one on the way. So, what
they internalized is the conversion rules 26 rather than the rules 27 of a native
speaker. A bilingual speaker would have both (25) and (27).
I shall now look at a special predicate, namely the truth predicate. Recall
that the predicate “It is true that”, formally “T”, is semantically rather vacuous.
For if ϕ is true, so is T(ϕ), and conversely. Moreover, as Tarski has pointed out,
the truth predicate is a dangerous device that leads to inconsistency. So what
to do?
Consider the sentence
It is true that it is raining. (28)
What does it take to understand this? The obvious proposal is to say that “ϕ
is true” is nothing but “ ϕ”. Thus, the formula describes a judgement (or a
disposition, whichever). Notice, however, that then we cannot represent the
idea that (28) is true. This would require writing
  It is raining. (29)

12 This may sound dangerous and confused. How, for example, are we to understand the expression

“man(x)” if English has no such syntax? And what is the point of introducing internal correlates?
My answer is this. First, “man(x)” is a piece of notation saying that the word /man/ is judged correct
for x. If you dislike my notation, choose another. Second, there is no need to introduce an internal
correlate. The mind can (and often does) take the words as standing in for themselves. There
simply are two roads to the concept: either by having it in the first place (under whatever name)
and then connecting a word to it, or by getting the word and learning to use it in accordance to the
concept without adding a different name to it.
Gnosis 415

However, “” cannot be used that way. The judgement sign attaches only to a
formula, not to a judgement. What it technically says is “there is a disposition
to consent to there being a disposition to consent to that it is raining” or
something of that sort. It is a second order notion. I claim that it does not
exist inside my head. And this is because the disposition of myself to consent to
some formula is not what I can apprehend. Recall that apprehension requires a
syntactic object, a formula. Facts are not apprehended. You don’t judge seeing
something; you only judge seeing something as something. Thus all I can do is
consent to “ϕ” and see myself doing that. The disposition has to be enacted in
order to become visible. Once I have observed myself giving consent to “ϕ” I
can express that in the thought: “ϕ is true”, or formally “T(ϕ)”, with “T” the
exponent of my own truth predicate. Thus, the rule of judgement for “T(ϕ)” is
this: judge “T( p)” true if ϕ. We get the following rule of judgement conversion.

 ϕ  T(ϕ) (30)

Notice that this can be iterated:

T S ϕ
T S  T(ϕ)
T S T(ϕ) ∅
(31)
T S T(ϕ)  T(ϕ)
T S T(ϕ)  T(T(ϕ))
T S T(ϕ) T(T(ϕ)) ∅

Namely, if I consent to “ϕ”, I am entitled to consent to “T(ϕ)”; once that

is done I can now apprehend “T(T(ϕ))”. Since “T(ϕ)” is there (on the slate,
representing my memory), I judge “T(T(ϕ))” to be true. What is crucial is to
understand that “is true” is a concept and “T” is a sign. Every layer of “T”
buries the content of the thought one level down; my original consent to “ϕ” is
history. Just now I have given my consent to “T(T(ϕ))”. The fact that this was
because I gave consent to “ϕ” is something I may recall from memory. It is not
something that I at this very moment can apprehend and judge. I can however
always return and rethink my judgement, but no two judgements can be made
at the same time.
At this point we may understand why paradoxes are not such a big concern
for speakers of natural languages. Suppose I meet the following inscription in
a classroom.

This sentence is false. (32)

In order to understand it I do not simply translate it into a formula and then

get a truth value in a model (which is impossible). Instead, the approach is in
stages. I might say: so this is allegedly true, let’s see. I convert /is false/ to
the disposition to reject the content, which is that very sentence. So I reject it.
I enter again, converting the meaning of /is false/ into the disposition to
accept, and so on. I may continue like a moth spiralling into the light, or else
416 M. Kracht

recall that I had reached that point before. I smile and leave. I refuse to do any
more work on that.
The logician in me might protest, thinking: how can the same thing be both
true and false? And how come you didn’t see it coming? Here I wish to answer
only the second complaint: because understanding is an act that unfolds in
time. It is an act that we may also refuse to perform or put to its proper
(?) conclusion. Normally, facts radiate to some degree. Our mind produces
conclusions in an instant. The word /Berlin/ invokes images in me that the
word and its meaning do not support; they are real for me, I have lived there
long enough for them to be automatic. But the radiation only goes a certain
way; I do not immediately start to picture everything I know about it, only a
little bit. And the same for the sentence above. The words it has in it typically
do not radiate very much. Since we have no intuitions about the sentence at all,
we go the pedestrian’s way, converting the words into representations, until we
either wake up to the fact that we have been fooled, or give up without result
at some point.13

10 Gnosis and Phatic Contour

I shall briefly turn to the connection with linguistics. This will be dealt with in
more detail in the sequel to this paper, [13]. A few things however are helpful
in seeing the relevance of the entire discussion. The claim I am advancing is
that linguistic meaning is not exclusively denotational in the standard sense; it
also encodes a kind of process meaning for the sentence.
Let me call an act of judgement a noetic act. Understanding, or gnosis,
is a sequence of noetic and other acts (for example, scheduling the rule
applications). Phatic acts are by contrast the public articulations of judgement.
They should not be confused with speech acts nor with noetic acts. Noetic acts
happen in the mind, phatic acts are acts of public expression. The crucial fact
now is that a judgement can be arrived at in different ways. What appears to
be logically identical may nevertheless be distinct at the level of gnosis. I give
a simple example. The sentence “John is tall.” may be seen as true simpliciter.
 tall ( j) (33)
We may however arrive at it in two steps. We picture John, and then judge that
he is tall. This is a two step process. This process can be recaptured symbolically
as a conditional judgement.
x = j  tall (x) (34)
(34) can be part of our judgement dispositions.
Essentially, I propose that sentence meanings do not take the form of
propositions but rather that of judgement dispositions. Thus, I claim that

13 See [3] on this point.

Gnosis 417

(33) and (34) are genuine representations. The division into assumptions and
judgements I call the phatic contour of that sentence. Notice that the phatic
contour is not to be confused with a sequence of phatic acts. If it were, (34)
would then be short for two phatic acts:
Œ Suppose: “x is John” is true.
 Then: “x is tall” is true.
But just as enactment of dispositions into a sequence of acts is not the same
as the disposition itself, so the phatic contour is different from the sequence
of phatic acts. In my view it is inappropriate to analyse (34) as containing a
hidden imperative (“suppose”) or a separate act of supposition. If you wanted
to express that sequence you’d better say something like the following.
Consider John. He is tall. (35)
Rather, in my view, (34) contains a single act, that of stating tallness of x, but
it is conditional on x being John. Thus there is a single phatic contour for
the entire sentence. The phatic contour is determined by the topic and focus
articulation or, similarly, the division into theme and rheme (see [17]). Given
the contour (34), /John/ is the theme, /is tall/ is the rheme. Note that I am
not concerned here with the linguistic encoding of the phatic contour. This will
be left for another occasion. I am interested here in demonstrating that it can
be motivated from the gnostic process.
The phatic contour contains a third element, the pheme (see [21]). It
expresses the phatic dimension. For example, “John is not tall.” may be given
the contour
x = j  tall (x) (36)
And “John might be tall.” may be given the contour
x = j  tall (x) (37)
Notice that in addition to asserting the falsity of “John is tall.”, the sentence
“John is not tall.” can also be used as a rejection of the judgement, that is, to
declare that the judgement fails. The contour would thus be14
x = j  tall (x) (38)
If the phatic contour were factually denotational it would say something about
the actual occurrence or the like of judgement acts. But that seems to me a
wrong way of putting things. Rather, I consider the phatic contour an offer
on the part of the speaker: speaker proposes a way to assess the logical (truth
conditional) content of his message. Hearer may comply but might be forced to
abandon this and perform his own gnosis. The crux is now this. As we have seen
earlier, different phatic contours can make a difference in processing effort;

14 Once again, note that “” denotes the act of coming to a negative conclusion, that is, rejecting

the judgement.
418 M. Kracht

some ways of saying the same are simpler than others, though we have also
seen that that depends on the state of the listener.

11 Variations

It is worthwhile to add a few remarks on the idea behind this calculus. To

some it may seem that I am advocating a return to psychologism in logic.15
This is not the case, though I admit that the matter is more complex. I certainly
do not agree that logical validities depend on someone’s psychological state. I
think that behind the use of logical symbols or words lies the intention on part
of the speaker to communicate an intersubjective meaning. Yet access to this
meaning is only made possible through calculations, and these are by necessity
private. However, if two persons share the language they will assign the same
meanings to the expressions unless they make errors in the computation.16
The reason why we end up with intuitionistic logic and not classical logic
is due to the interpretation of the implication. Although we may conceive
of implication as being defined via truth tables, the definitions given of
“if· · · then” in English suggest a proof theoretic reading in the way given here.
All else depends on the calculus of derivations. As I have portrayed matters
the calculus seems to be fixed once and for all. That is however not the way I
see it. I see the calculus here as a basic formalism. In the long run anyone may
come to acquire new (or different) modes of inference (such as conversion
rules). We can, for example, opt for the following rule.
T, S,  ¬ϕ  T, S,  ϕ (39)
Once we add this rule we get classical logic, as the formula ¬¬ϕ → ϕ now
becomes derivable.
There is nothing to rule this out in principle. And I do not see why there
should be anything that rules this out. For the intended meaning of negation
words like “not” may indeed be that of non-truth. And if that is so then the
rule above makes perfect sense. It is likewise conceivable that someone thinks
of Pierce’s formula as an eternal truth and makes it an axiom, though I find
that highly unlikely.
Another problematic point may be the treatment of propositional attitudes.
At the surface it looks as if I am subscribing to a view that sees propositional
attitudes (inasmuch as they are modes of judgement) as relations between
an individual and a syntactic object. This view faces many problems which I
cannot discuss here (see [10]). However, it is not necessary within the present
theory to take that stance. One reason is that the notion of syntactic object is
sufficiently flexible: it does not have to be an object of an external language
(English, Hungarian or whatever) and it can in principle be an object of

15 Thanks to the reviewer for pressing this issue.

16 Or the use different calculi, see the discussion below.
Gnosis 419

an internal language or a proposition (which at least [10] equates with a

syntactic object, namely L(ogical) F(orm)). The other reason is that there is a
distinction between an immediate attitude and a mediated attitude, parallel to
the distinction between explicit and implicit knowledge in [7]. I have a direct
belief that unicorns do not exist if in the belief mode I consent to the object
expressing this content. I have a mediated belief of that fact if I come to consent
to this object (in the belief mode) after some reasoning steps. Mediated beliefs
are closed under consequence, direct beliefs are not.

12 Conclusion

The present paper has tried to develop a somewhat more realistic model of
reasoning. It is based on the idea that reasoners can pay attention only to one
formula at a time. When asked whether a formula is true they either know the
answer off hand or have to decompose it to see what is inside. If its structure
conforms with the unpacking mechanism they can arrive at an answer. If not,
difficulties arise that they may not effectively be able to resolve.
One result of this paper is that the logic resulting from this process oriented
view on logic is intuitionistic logic and not classical logic. In this connection it is
important to emphasise that the rule system above is only a proposal, nothing is
sacrosanct about it. Moreover, since it makes empirical predictions about ease
of understanding, it would be desirable to check the predictions empirically.
Though there is some latitude, it seems to me quite a robust property that the
calculus derives intuitionistic logic, not classical logic.

References

1. Barsalou, L. W. (2008). Grounded cognition. Annual Review of Psychology, 59, 617–645.

2. Brentano, F. (1933). Kategorienlehre. Hamburg: Felix Meiner Verlag.
3. Frege, G. (1983). Logik in der Mathematik. In H. Hermes, F. Kambartel, & F. Kaulbach (Eds.),
Gottlob Frege: Nachgelassene Schriften (2nd ed.). Felix Meiner Verlag.
4. Gaerdenfors, P. (1988). Knowledge in Flux. MIT Press.
5. Gillies, A. (2004). Epistemic conditionals and conditional epistemics. Nôus, 38, 585–615.
6. Groenendijk, J., & Stokhof, M. (1991). Dynamic predicate logic. Linguistics and Philosophy,
14, 39–100.
7. Hintikka, J. (1962). Knowledge and belief. An introduction into the logic of the two notions.
Ithaca: Cornell University Press.
8. Husserl, E. (1975). V. (Fünfte) Logische Untersuchung. Hamburg: Felix Meiner Verlag.
9. Kearns, J. T. (1997). Propositional logic of supposition and assertion. Notre Dame Journal of
Formal Logic, 38, 325–349.
10. King, J. C. (2007). The nature and structure of content. Oxford: Oxford University Press.
11. Kracht, M. (1990). Traditional linguistics can solve logical puzzles. Preprint 12/90 of the
Gruppe Logik, Wissenstheorie und Information, FU Berlin.
12. Kracht, M. (1992). Assertivity, theme and presupposition. Preprint 19/92 of the Gruppe Logik,
Wissenstheorie und Information, FU Berlin.
13. Kracht, M. (2010). The inner dialog: Pragmatics for one person. In K. Robering, & S. Bab
(Eds.), Judgements and propositions. Logische Philosophie (number 21). Logos Verlag.
420 M. Kracht

14. Moschovakis, Y. (1994). Sense and denotation as algorithm and value. In J. Oikkonen, &
J. Väänänen (Eds.), Proceedings of the ASL meeting 1990, Helsinki. Lecture Notes in Logic
(number 2, pp. 210–249). Berlin and Heidelberg: Springer.
15. Putnam, H. (1975). The meaning of ‘meaning’. In Mind, language and reality (pp. 215–271).
Cambridge: Cambridge University Press.
16. Rautenberg, W. (1979). Klassische und nichtklassische Aussagenlogik. Braunschweig/
Wiesbaden: Vieweg Verlag.
17. Vallduví, E. (1990). The information component. PhD thesis, University of Pennsylvania.
18. van Lambalgen, M., & Hamm, F. (2005). The proper treatment of events. In Explorations in
Semantics (number 4). Oxford: Blackwell.
19. Veltman, F. (1985). Logics for conditionals. PhD thesis, Department of Philosophy, University
of Amsterdam.
20. Vermeulen, K. (2000). Text structure and proof structure. Journal of Logic, Language and
Information, 273–311.
21. Zemb, J. (1978). Vergleichende Grammatik Französisch-Deutsch. Comparaison de deux sys-
tèmes (Vol. 1). Mannheim: Bibliographisches Institut.