See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/265332203

Information Dynamics in Cognitive, Psychological, Social and Anomalous

Phenomena

Book · January 2004

DOI: 10.1007/978-94-017-0479-3

CITATIONS READS

258 257

1 author:

Andrei Khrennikov
Linnaeus University
1,139 PUBLICATIONS   17,794 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

4th International Electronic Conference on Entropy and Its Applications View project

Analysis on adele ring View project

All content following this page was uploaded by Andrei Khrennikov on 05 April 2019.

The user has requested enhancement of the downloaded file.

 [Information Dynamics] Information Dynamics in
Cognitive, Psychological,
Social, and Anomalous
Phenomena

1
 International Center for Mathematical Modeling
in physics, engineering, and cognitive science
Växjö University, Sweden
*Contents Contents
List of Figures
ix Introduction xiii
1.
PROCESSING OF INFORMATION ON P -ADIC TREES1 1.1Ultrametric
Spaces 4

1.2m-adic Geometry 6

1.3Geometry of Information Spaces 8

1.4Dynamical Processing of Information 10

1.5Role of Hierarchical Structure 13

1.6Role of Chance in Processing of Cognitive Information 17

1.7Information Reductionism 19
2.
2.1Hierarchical Coding of
HIERARCHY OF INFORMATION23
Information 26

2.2Flows of Associations, and Ideas 32

2.3How Can the Brain Play Dice? 34

2.4Constraints on Information Spaces 37

3.
P -ADIC DYNAMICAL SYSTEMS39 3.1p-Adic Numbers 41

3.2Roots of Unity 43

3.3Dynamical Systems in Non-Archimedean Fields 44

3.4Dynamical Systems in the Field of Complex p-adic Numbers 47

3.5Dynamical Systems in the Fields of p-adic Numbers 48

3.6p-adic Ergodicity 50
3.7Newton’s Method (Hensel’s Lemma) 54

3.8Computer Calculations for Fuzzy Cycles 55

4.
RANDOM PROCESSING OF INFORMATION57 4.1Random Dy-
namical Systems 58

4.2Longterm Behaviour, Dynamics on the Attractor, Examples 65

4.3Consequences for Cognitive Sciences 67

5.
INFORMATION QUANTUM MECHANICS71 5.1Quantum-Like
Formalism for a One-Layer Brain 74

5.2Motivation Observable 76

5.3Neuron Activation Observable 79

5.4Complex Cognitive Systems; Evolution 81

6.
6.1Newton
BOHMIAN MECHANICS ON INFORMATION SPACES85
Laws for Information Processes 87

6.2Bohmian Mechanics for Hierarchical Information 89

6.3Interactions between Information Systems 91

6.4Hamiltonian Equations and Active Information 93

6.5Information Mass 95

6.6Wave Functions Taking Values in p-adic Fields 96

6.7Information Waves on p-adic Trees 100

6.8 p-adic Bohmian Mechanics and Waves of Brain Activation 103

6.9Conservation Laws 108

6.10
Mechanics of a System of Information Transformers, Con-
straints on Information Spaces 110

2
6.11
Social and Anomalous Phenomena 114
7.
7.1Abstract
ABSTRACT ULTRAMETRIC INFORMATION SPACES117
Ultrametric Spaces 117

7.2 Hierarchy of Associations 122

7.3Topology and Materialism 122

7.4Existence of Universal Mental Space 123

7.5Towers of Associations 125

7.6Infinite Information Towers 126

8.
PATHWAY REPRESENTATION OF COGNITIVE INFORMA-
TION1318.1Model: Thinking on a Cognitive Tree 133

8.2Dynamics in the Information Space 138

8.3Diffusion Model for Dynamics of Mental State 141

8.4Information Phase Space 143

8.5Mental State as the Distribution of a p-adic Random Walk 144

8.6Discussion of the Neural Pathways Thinking Model 147

9.
CONTEXTUAL APPROACH TO QUANTUM THEORY153 9.1The
Växjö Interpretation of Quantum Mechanics 157

9.2Contextual Viewpoint of Quantum Stochastics 165

9.3Law of Statistical Balance in Nature 175

9.4Experiments on Quantum-Like Behaviour of the Mind 182

9.5Experimental Confirmation 184

10.
FREQUENCY ANALYSIS OF FOUNDATIONS OF QUANTUM
10.1
MECHANICS187 Classification of Transformations of Proba-
bility 187

3
10.2
Classical, Quantum, and Non-Classical-Quantum Physics 192

10.3
Hyperbolic Probabilistic Behaviour 195

10.4
Linear Space Representation of the Trigonometric Rule 197

10.5
Linear Space Representation of the Hyperbolic Rule 200

10.6Conclusions 204
11.
BOHMIAN MECHANICS FOR FINANCIAL PROCESSES209
11.1
Price Phase Space 211

11.2
Hamiltonian Price Dynamics and the Stock Market 215

11.3
Financial Pilot Waves 215

11.4
The Dynamics of Prices Guided by the Financial Pilot Wave 217
References
224 Index
233

List of Figures
1 The 2-adic tree . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2 The factorial tree ZM for M = (2, 3, 4, . . .). . . . . . . . . . . . 15
3 The Markov chain given by φ on I(29,2,3) , (q1 is omitted). . . . 73
4 Lattice of balls for the ultrametric space X . . . . . . . . . . . 127
5 Lattice of balls for the ultrametric space Z2 . . . . . . . . . . 127
6 Tree-representation of the maxmetric space for N natural num-
bers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7 Images of associations . . . . . . . . . . . . . . . . . . . . . . 136
8 Idea V3 (u3 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
9 The tree of the space of ideas . . . . . . . . . . . . . . . . . . 247
10 The tree of the space of ‘super-ideas’ . . . . . . . . . . . . . . 247
11 Centered pathway . . . . . . . . . . . . . . . . . . . . . . . . . 247
12 A tree of S-centered pathways: a ‘cognitive tree’ . . . . . . . . 247
13 cognitive tree — a . . . . . . . . . . . . . . . . . . . . . . . . 248
14 cognitive tree — b . . . . . . . . . . . . . . . . . . . . . . . . 248
I dedicate this book to Mother and Father ă ă ă ă
ă ă In this book we develop various mathematical models of information
dynamics, I-dynamics (including the process of thinking), based on methods

4
of classical and quantum physics. The main aim of our investigations is to
describe mathematically the phenomenon of consciousness. We would like
to realize a kind of Newton–Descartes program (corrected by the lessons
of statistical and quantum mechanics) for information processes. Starting
from the ideas of Newton and Descartes, in physics there was developed
an adequate description of the dynamics of material systems. We would
like to develop an analogous mathematical formalism for information and,
in particular, mental processes. At the beginning of the 21st century it is
clear that it would be impossible to create a deterministic model for general
information processes. A deterministic model has to be completed by a
corresponding statistical model of information flows and, in particular, flows
of minds. It might be that such an information statistical model should have
a quantum-like structure.
ă Human thinking (as well as many other information processes) is fun-
damentally a hierarchical process. This hierarchical internal structure of
information processes in the brain is the basis of mental holicity. One of
the exhibitions of mental holicity is the impossibility of localizing mental
(in particular, psychological) functions, see, for example, A. Damasio [52],
J. Fuster [81] for an extended bibliography in neurophysiology of this prob-
lem. It must be underlined that localization is localization with respect to
some geometry. The standard interpretation of localization in neurophysiol-
ogy and cognitive science is localization with respect to Euclidean geometry.
And neurophysiological and cognitive experience tells us such an Euclidean
localization is impossible. In this book we demonstrate that if, instead of
Euclidean geometry, we consider a special hierarchical geometry, so called
m-adic geometry, then we obtain a model in which mental (in particular,
psychological) functions are well localized.
ă One of the main problems is to find adequate mathematical models for
information spaces and, in particular, mental spaces. The standard math-
ematical model of space (which is used in classical as well as in quantum
physics) is constructed on the basis of three straight lines. In the mathemat-
ical model each line is described by the set of real numbers R. In the simplest
case the physical space is represented as the direct product (Descartes space)
R × R × R of three straight lines. In advanced models we use ‘curved trans-
formations’ of real lines as the basis of the physical space. In the theory
of relativity we use four real lines instead of three (see also the theory of
strings [85] for multidimensional, for example 26-dimensional, real spaces).
We remark that an essential part of physical investigations during the last
two centuries has been devoted to the development of new (increasingly ad-
vanced) models for physical space. However, practically all models used in
physics are still models based on the real line (or its curved transformations).

5
From this point of view the appearance of non-Euclidean geometric mod-
els (starting with the famous works of Lobachevsky) did not imply crucial
changes in physics. We still work in modified variants of the space which was
used by Newton. Of course, these real spaces provided fruitful descriptions
of various natural phenomena. However, I think that we could not provide
a description of the whole of physical reality by using the physical spaces of
the 18th–20th centuries. In particular, cognitive phenomena and, especially,
the phenomenon of consciousness could not be described on the basis of old
mathematical models of space. The human mind could not be embedded
into the real line. Even spaces constructed on the basis of curved real lines
could not contain the mind’s trajectory. A curved real line is not curved suf-
ficiently to describe the flow of a mind. ă We use a new mathematical model
of space (and space–time) to describe the flow of a mind. Each coordinate of
this space has a treelike structure; each branch of this tree has a hierarchical
structure.
ă In fact, we use a rather restricted class of trees. These are so called
m-adic trees. Here m > 1 is a natural number which describes the homo-
geneous structure of the tree: m branches leave each vertex of the tree. In
physical models (for purely mathematical reasons) we consider the special
class of these trees in which m = p is a prime number. This gives the pos-
sibility of introducing the arithmetical operations of addition, subtraction,
multiplication, and division between branches of p-adic trees. In our models
this p-adic arithmetic is realized as ‘mind arithmetic’. We remark that there
exists well developed mathematical analysis on p-adic trees [163]. By using
p-adic analysis we consider a few classical (neural network) and quantum
(including Bohmian mechanics) models of I-dynamics on p- adic hierarchical
trees with applications to psychology, neurophysiology, social, and cognitive
sciences.
ă We remark that p-adic numbers (which give the numerical represen-
tation of the p-adic tree) have been used intensively in physics since the
1980s. Investigations of p-adic physical models were started at the Steklov
Mathematical institute (Moscow): Vladimirov, Volovich, Khrennikov, Ze-
lenov,..., see, for example, [187], [107], [111]. The main idea of Vladimirov
and Volovich was that the geometry of space–time at the ‘deepest levels of
the micro-world’, at so called Planck distances, could not be the same as
in macro-world. They speculated that, instead of the real line, we have to
use a p-adic tree to describe space–time coordinates of Planck sized objects.
This idea was intensively exploited in string theory (Vladimirov, Volovich,
Freund, Witten, Parisi, Aref’eva, Frampton, Dragovich, ...,) which describes
fundamental micro-objects, so called strings. ă In fact, I became interested
in applications of p-adic numbers through my contacts with p-adic physicists.

6
However, I would not like to give the impression that there must be a direct
relation between the use of p-adic numbers in micro-physics and our p-adic
modeling of cognitive phenomena. ă In our information modeling the main
distinguishing feature of p-adic numbers is the treelike hierarchical structure.
In fact, this book is devoted to classical and quantum models of flows of
hierarchically ordered information.
ă

ă Växjö, Bonn, Havana, 1999–2002. ă

[Processing of Information on p-adic Trees] Processing of Information
on p-adic Trees
The notion of informationinformation will play an important role in our
book. We do not follow Shannon’s theory [169] that there is a quantative
measure of information which represents the way in which the state of a sys-
tem is uncertain for us. Shannon’s information can be called passive infor-
mationpassive information. Following D. Bohm and B. Hiley [37] we consider
information as active informationactive information. An information space
XI is nothing other than a set of strings of symbols with some operations
on strings. Such operations define interactions between strings, the active
exchange of information.
Everywhere in this book we shall use the abbreviation I for the word
‘information’. The symbol τ will be used to denote ‘cognitive systems’.
Cognitive systems (in particular, human beings) operate with I-strings.
Such strings of information (just sequences of digits) will be called I-states.
Hierarchically coupled families of I-states form new I-objects. They will be
called associations. Families of associations form I-objects of a higher level.
They will be called ‘ideas’.
This chapter is devoted to mathematical models of dynamical thinking at
the level of the simplest I-structures, namely I-states. Models of dynamical
thinking at higher levels of I-hierarchy, namely associations and ideas, will
be studied in chapter 2. Of course, the latter models are more interesting
for cognitive sciences, because they give the possibility of describing flows of
highly organized information. However, even ‘primitive thinking dynamics’
at the level of I-states describes some essential aspects of human psychology.
The process of thinking (at the primary level) is performed by a dynamical
system which works with I-states, i.e., there is a nonlinear relation between
input and output I-states,
xn+1 = f (xn ), xn ∈ XI , (1)
where XI is the configuration space of the dynamical system – the space
of I-states, or I-space. To use dynamical systems (feedback processes) for

7
the investigation of the functioning of the human brain is quite natural. The
main problem is to construct a good mathematical model which describes the
essential aspects of this functioning. The first problem is to find an adequate
mathematical description of I-states x. The second problem is to find the
corresponding dynamical laws f (x).
In this book we propose a mathematical model for the process of thinking
which is based on hierarchical coding systemshierarchical coding systems.
These coding systems are described by systems of so called p-adic, or more
general m-adic, numbers. Here m > 1 is a natural number, the basis of the
coding system. From the mathematical viewpoint the model is essentially
simpler for m = p, where p is a prime number (for example, p = 2). Therefore
we often restrict our considerations to this case.
Another basic feature of our model is that the process of thinking is split
into two separate (but at the same time closely connected) domains: the con-
sciousconscious and the unconsciousunconsciousOf course, the author was in-
fluenced by Freud’s theory about the unconscious and its role in the process
of thinking, [78].. We shall use the following point of view of the simultaneous
work of the consciousness and unconsciousness. The consciousness contains a
control center CC which has functions of control. It formulates problems and
sends them to the unconscious domain. The process of finding a solution is
hidden in the unconscious domain. In the unconscious domain there works a
gigantic dynamical system. Its work starts with an I-state x0 (or a group of
I-states U0 ) which has been communicated by the consciousness. Mathemat-
ically it corresponds to the choice of an initial point x0 (or a neighborhood
U0 ). Starting with this initial point x0 a thinking processorthinking processor
π in the unconscious domain generates at tremendous speed a huge number
of new I-states:
x1 = f (x0 ), ...., xn+1 = f (xn ), ...
These I-states are not used by the consciousness. The consciousness (namely
CC ) controls only some exceptional moments in the work of the dynamical
system in the unconscious domain. These are different regimes of stabi-
lization. First, there are attractorsattractors which are considered by the
consciousness as possible solutions of the problem. Then there are cycles
(a → b → · · · → c → a) which generate signals to stop the work of the
dynamical system. If the consciousness cannot take a decision then it can
send a new initial I-state x00 to the unconscious domain or change the regime
of work of a thinking processor in the unconscious domain. Mathematically
the change of a regime can be described as the change of a function f (x)
which determines the dynamical system. Thus we can describe the process
of thinking as the work of a family of dynamical systems fα (x), where the pa-

8
rameter α is controlled by the consciousness (or chance in random dynamical
thinking models).
Geometrically we can imagine a system of m-adic integers (which will be
the mathematical basis of our cognitive models) as a homogeneous treetree
with m-branches splitting at each vertex. This tree starts with the symbol ?.
This is the root of the m-adic tree. There exists an m-adic algebraic structure
on this tree which gives the possibility of adding, subtracting and multiplying
branches of this tree. I-states of cognitive systems will be represented as
branches of such a tree. We have an algebra of informationWe do not claim
that the real process of thinking is reduced to an m-adic algebraic dynamics.
Such an ‘algebraic thinking’ might be useful for cognitive simulation.. At the
brain’s ‘hardware’ level such branches may be represented by hierarchical
chains of neuronshierarchical chains of neurons, see chapter 2. The distance
between I-states is determined by the length of their common root: close
I-states have a long common root. We remark that this m-adic metric is
well known in number theory. The corresponding geometry strongly differs
from ordinary Euclidean geometry.
Cognitive I-dynamics is not only the dynamics of I-states (branches
of a tree). Families of I-states also have a cognitive meaning. In our
model bundles of branches of a hierarchical tree have the meaning of associa-
tionsassociations. An association is a family of I-states which are sufficiently
close in the m-adic metric. In our model families of associations are inter-
preted as ideas. Our main aim is to the study dynamics of associations and
ideas, namely the collective dynamics of I-states, see chapter 2.

0m

:
:
 XXX
 z
0mX
 
*
 XX XXX
1m
 z 
:
 XXX
z

?m
HH
0m
HH 
:
:
 XXX
HH  z
1mX
j  
XX XXX
1m
z 
:
XXX
z

Figure 1: The 2-adic tree

9
1 Ultrametric Spaces
The notion of a metric space is used in many applications for describing
distances between objects.
Let X be a set. A function ρ : X × X → R+ (where R+ is the set of
positive real numbers) is said to be a metricmetric if it has the following
properties:
1)ρ(x, y) = 0 iff x = y (non-degenerated);
2)ρ(x, y) = ρ(y, x) (symmetric);
3)ρ(x, y) ≤ ρ(x, z) + ρ(z, y) (the triangle inequality).
The pair (X, ρ) is called a metric space.
Abstract metric spaces were introduced as generalizations of the Eu-
clidean space
Rn = {x = (x1 , ..., xn ) : xj ∈ R}
with the standard metric
p
ρE (x, y) = (x1 − y1 )2 + ... + (xn − yn )2 .

However, in some applications the point structure of X and the properties

of a metric ρ may essentially differ from the Euclidean case.
We are interested in the following class of metric spaces (X, ρ). Every
point x has an infinite number of coordinates

x = (α1 , ..., αn , ...) . (2)

Each coordinate yields the finite number of values

α ∈ Am = {0, ..., m − 1}, (3)

where m > 1 is a natural number, the base of the alphabet Am . The metric
ρ must be the so called ultrametric [127], [72], [163], i.e., satisfy the strong
triangle inequalitystrong triangle inequality:

ρ(x, y) ≤ max[ρ(x, z), ρ(z, y)], x, y, z ∈ X. (4)

The strong triangle inequality can be stated geometrically: each side of a

triangle is at most as long as the longest of the two other sides. It is impos-
sible to imagine such a ‘triangle’ in ordinary Euclidean space. However, we
shall see that such triangles of associations are natural objects in I-spaces.
We note that the strong triangle inequality implies the ordinary triangle
inequality.

10
 We denote the space of sequences (2) by the symbol Zm . The standard
ultrametric is introduced on this set in the following way.
Let us fix a real number 0 < q < 1. Let
x = (α0 , α1 , α2 , ...., αn , .....), y = (β0 , β1 , β2 , ..., βn , ...) ∈ Zm .
We set

ρm (x, y) = q k if αj = βj , j = 0, 1, ..., k − 1, and αk 6= βk .

This is a metric and even an ultrametric. To find the distance ρm (x, y)

between two strings of digits x and y we have to find the first position k at
which the strings have different digits. The choice of the constant q does not
play any role. Geometries (topologies) corresponding to different 0 < q < 1
are equivalent. The standard choice is q = m1 . Thus

1
ρm (x, y) = .
mk
Let m = 2. Let x = (0, 1, 0, ...) and y = (0, 1, 1, ...). Here k = 2 and,
hence, ρ2 (x, y) = 1/4.
Example 1.1. Let A = (a0 , ..., am−1 ) be an alphabet having m letters.
The set of all (infinitely long) texts in this alphabet can be identified with
Zm . Every text x can be described by a sequence

x = (α0 , ..., αk , ...), αj ∈ A. (5)

We use the letter a0 for a blank. In such an alphabet a finite text can be
identified with a sequence x in which αj = a0 for all j ≥ k, k = k(x).
Let A be the English alphabet including the blank a0 , the dot, and all
other grammatical signs. Let us consider the texts
x =(I have to go to the University),
y = (I go to the University), z = (I go to a shop).
Then ρm (x, y) = ρ(x, z) = m12 (the common initial segment of length 2);
ρm (y, z) = m18 (the common initial segment of length 8). In particular,
1
m2
= ρm (x, y) ≤ max[ρm (x, z), ρm (z, y)] = m12 .
We now consider the text w = (He has to go to the University). Then
ρm (x, w) = ρm (y, w) = ρm (z, w) = 1. This illustrates the important property
of our model: the choice of the first bits of a text plays a crucial role (points
which begin with ‘I’ are strongly separated from points which begin with
‘he’ or ‘she’).
Let (X, ρ) be an arbitrary ultrametric space. For r ∈ R+ , a ∈ X, we set

Ur (a) = {x ∈ X : ρ(x, a) ≤ r}, Ur− (a) = {x ∈ X : ρ(x, a) < r}.

11
These are ballsball of radius r with center a. Balls have the following prop-
erties [163]:
1) Let U and V be two balls in X. Then there are only two possibilities:
(a) balls are ordered by inclusion (i.e., U ⊂ V or V ⊂ U ); (b) balls are
disjointThere is a third possibility in Euclidean space..
2) Each point of a ball may serve as a centre.
3) In some ultrametric spaces a ball may have infinitely many radii.
The reader who knows a little bit of topology may be interested in the
following property of balls:
4) Each ball in X is both open and closed.
Sets which are open and closed at the same time will play a large role in
our further investigations. We use the word clopen as the abbreviation for
closed and openIn Euclidean space each ball Ur− (a) is open but not closed;
each ball Ur (a) is closed but not open..
The symbol Sr (a) denotes the spheresphere

Sr (a) = {x ∈ X : ρ(x, a) = r}

of radius r ∈ R+ with centre a.

2 m-adic Geometry
The system of p-adic numbers Qp (p > 1 a prime number) was constructed
by K. Hensel [91], see chapter 3 for the details. During the last years these
numbers have been intensively used in theoretical physics (see, for example,
the books [187], [107], [111] and papers [190], [186], [79], [80], [105], [109],
[110], [13], [14], [2], [116]), the theory of probability [117], [106], [112], inves-
tigations in chaos and dynamical systems [182], [16], [17], [136], [95], [108],
[111], [3]–[6], [113]–[116], [119]—[122], [131], [145].
To use only number systems which are constructed on the basis of prime
numbers p is not natural from the biological point of viewHowever, models
based on a prime number p are simpler from the mathematical point of view.
Therefore we shall study only p-adic models at the mathematical level of
rigorousness.. Therefore we start with more general number systems, m-adic
numbers, where m > 1 is a natural number. Then we shall explain why
mathematicians prefer to use p-adic numbers, see chapter 3 for the details.
In chapter 3 we shall introduce systems of p-adic numbers in number theo-
retical framework. Such a number theoretical approach is based on advanced
mathematical apparatus. At the moment we prefer to work by using only
simplest mathematical tools. We would like to split difficulties in modelling
of information processes and mathematical difficulties.

12
 In the present chapter we introduce p-adic and general m-adic, where m
can be non-prime, number systems by using the metric space approach.
Let m > 1 be the fixed natural number. We consider the metric space
(Zm , ρm ). In fact, this metric space has the natural algebraic structure, see
Hensel [91] and chapter 3.
First, we remark that this metric can be presented in the form ρm (x, y)
= |x − y|m , where | · |m is the so called m-adic valuation (an analog of
the ordinary real absolute value). It is defined in the following way. Let
x = (α0 , α1 , ..., αn , ...). Then
1
|x|m = if αj = 0, j = 0, ..., k − 1, and αk 6= 0 .
mk
We remark that the strong triangle inequality holds for the m-adic valuation:

|x + y|m ≤ max[|x|m , |y|m ].

The strong triangle inequality is the geometric cornerstone of our cogni-

tive models. We think that the topologies on spaces of ideas differ crucially
from topologies on ‘physical spaces’, [43], [34], [56], [64], [89], [191], [32], [33],
[188]. The latter spaces are equipped with metric topologies which are not ul-
trametric (i.e., a ‘physical metric’ satisfies only the ordinary triangle inequal-
ity, but it does not satisfy the strong triangle inequality). Relations between
ordinary metric topologies of ‘physical spaces’ and ultrametric topologies of
cognitive spaces will be studied in chapter 7.
We remark that (as in the case of the real valuation, the standard absolute
value on the real line, | · |) :

|x|m ≥ 0 and |x|m = 0 iff x = 0.

A point x = (α0 , α1 , α2 , ..., αn , ....) of Zm can identified with the ‘number’:

x = α0 α1 ...αk .... ≡ α0 + α1 m + ... + αk mk + ... . (6)

(this series converges in the metric space Zm ). In particular, a finite I-string

x = α0 α1 ...αk can be identified with the natural number

x = α0 + α1 m + ... + αk mk .

Therefore the set of all finite I-strings can be identified with the set of nat-
ural numbers N. So the dynamics of finite I-strings can be simulated via
dynamics on N. Moreover, N is the dense subset of Zm : each x ∈ Zm can
be approximated with an arbitrary precision by natural numbers. If x is a

13
natural number then |x|m = m−k iff x is divisible by mk and not divisible by
mk+1 .
The set Zm is called the set of m-adic integers.
It is possible to introduce algebraic operations on the set of m-adic num-
bers Zm , namely addition, subtraction, and multiplication. These operations
are natural extensions of the standard operations on the set of natural num-
bers N = {0, 1, 2, 3, ...} (see chapter 3 for the details). We remark that
division is not well defined on Zm (it is a number ring, but not a number
field, [163]).
Let us consider expressions of the form:
∞
X
x= αj mj . (7)
j=s

where αj = 0, 1, ..., m − 1 and s = 0, ±1, ±2, ... Denote the set of all such
expressions by the symbol Qm . We set |x|m = m−s if αs 6= 0. This is the
natural extension of the valuation on Zm : if s = 0, 1, 2, ..., so that x =
αs ms +...+αj mj +..., then |x|m = m−s . On the other hand, if s = −1, −2, ...,
so that x = α−k /mk + ... + αj mj + ..., where k = −s, then |x|m = m−s = mk
Ps system jof real numbers R consists of expressions of the form: x =
The
j=−∞ αj m (the most useful is the decimal scale). In the real case there can
be only a finite number of terms with positive powers of m, and in the m-adic
case there can be an infinite number of terms with positive powers of m. The
situation is the opposite for negative powers.. Qm is a complete ultrametric
metric space with respect to the corresponding metric, ρm (x, y) = |x − y|m .
We introduce addition, subtraction, and multiplication in Qm in the same
way as in Zm by extending the standard operations on set of finite sums
n
X
xn = αj mj .
j=−s

We note that
|xy|m ≤ |x|m |y|m .
If m = p is a prime number then (as for the standard real absolute value)

|xy|p = |x|p |y|p .

In general, division is not well defined in Qm . However, if m = p is a prime

number (p = 2 or 3, or 5, or 1999,...), then division is well defined. We note
that the set Zm coincides with the unit ball U1 (0) of Qm :

Zm = {x ∈ Qm : |x|m ≤ 1}.

14
 In chapters 1–2 we shall use only the space Zm as the mathematical
model of I-space; in chapter 5–6 we shall also discuss the possibility to use
Qm . We can also use more complicated number systems corresponding to
non-homogeneous scales M = (m0 , m1 , ..., mk , ...), where mj > 1 are natural
numbers. Here the digits αj in (2) yield values 0, 1, ..., mj −1, i.e., an alphabet
depends on the place of a letter in a text. In this way we can construct the
number system QM . However, here the situation is quite complicated from
the mathematical point of view. Therefore we shall not use mathematical
models over QM .

0m
:

 *


: X
z
X
H
j
H
m  -
1m
*

0 XXX
 :


X
z
X
H
j
H

* XXX
 2m
z :

 *

 X
z
X
H
j
H

?m
HH
0m
H :

 *

HH :
 X
z
X
H
 j
H
1mXXX - 1m
j
H  :
 *


X
z
X
XXX H
j
H
2m
z :

 *

X
z
X
H
j
H

Figure 2: The factorial tree ZM for M = (2, 3, 4, . . .).

3 Geometry of Information Spaces

We shall use the following mathematical model for I-space:
(1) Set-structure: The set of I-states XI has the structure of the m-adic
tree: XI = Zm .
(2) Topology: Two I-states x and y are close if they have a sufficiently
long common root. This topology is described by the metric ρm .
So in our mathematical model of I-space is represented as the metric
space (Zm , ρm ).
To construct a mathematical dynamical model for thinking we have to
describe interactions between I-states. We represent these interactions by
using transformations of XI which generate a new I-state x = g(x1 , ..., xn )
on the basis of I-states x1 , ..., xn . The simplest interactions are realized as
binary operations. In particular, we can consider the algebraic operations on
the space of I-states XI = Zm .

15
 For example, consider the space of I-states Z2 . Let x = (010...0...), y =
(100...0...). Then z = x + y = (110...0...). Thus if the I-state x was started
with (no,yes,...) and the I-state y with (yes,no,...), then the addition interac-
tion between these I-states produces the I-state z starting with (yes,yes,...).
Let x = (100...0...), y = (110...0...). Then z = x+y = (0010...0....). The dou-
ble ‘yes’ at some level produces ‘no’ on this level and ‘yes’ on the next level.
The multiplication on Z2 induces the following interaction between I-states.
Let x and y be presented in Z2 by natural numbers, x = 2n , y = 2m (only
one ‘yes’, respectively, on the nth and mth levels). Then z = xy = 2n+m
(‘yes’ on the (n + m)th level). Thus the multiplication interaction generates
a jumping change of I-states. The problem of the mathematical description
of interaction functions is the hard problem in cognitive modeling. At the
present time we do not know the forms of transformations of hierarchical trees
that describe interactions between I-states (for example, in the human brain
or lion’s brain). In this book we consider some transformations by purely
mathematical arguments. However, we observe that these transformations
of an m-adic tree simulate some essential features of cognitive behaviour.
The last step in the construction of a mathematical model is the descrip-
tion of the evolution process in the space of I-states. We must determine a
function f : XI → XI which describes the dynamical system of unconscious-
ness. In this way we realize the process of thinking as a dynamical system
xn = f (xn−1 ) on the space of I-states XI . Starting with an I-state x0 we
obtain a chain of I-states x0 , x1 , ..., xn , ... . This is the process of thinking.
We are interested in continuous maps f : XI → XI . According to our defi-
nition of the nearness of I-states, f maps the I-states x and y which have a
long common root into the I-states x0 = f (x) and y 0 = f (y) which also have
a long common root. These ‘common root’-smoothness of dynamical laws
gives the possibility of lifting the dynamics of I-states to the dynamics of
associations (bundles of branches having common roots), see chapter 2, and
then ideas (families of bundles of branches).
Dynamical thinking is performed via the following procedure: an initial
I-state x0 is sent to the unconscious domain; it is iterated by some dynamical
system (determined by a map f : Zm → Zm ); an attractor is communicated
to the consciousness; this is the solution of a problem x0 . It may be that
iterations starting with some x0 will not arrive at any attractor. For example,
starting with x0 , τ may perform a cyclic behaviour in the process of thinking.
In such case a cognitive system τ would not find the definite solution of a
problem. We remark that it is impossible to escape a cyclic behaviour at
the level of I-states: even the simplest dynamical systems in Zm may have a
huge number of cycles of various lengths; see [122], [145].

16
4 Dynamical Processing of Information
As we have already mentioned, analysis in number systems Qm , m 6= p,
and QM is not yet developed. Therefore we shall consider cognitive models
described by p-adic numbers.
We shall study dynamical systems corresponding to maps:

Zp → Zp , x → f (x). (8)

As usual, we study the behaviour of iterations xn = f n (x0 ), x0 ∈ Zp , where

f n (x) = f ◦ ... ◦ f (x) = f (....(f (f (x)...),

the result of n successive applications of the map f. We shall use the standard
terminology of the theory of dynamical systems; see, for example, [25], [192].
If f (x0 ) = x0 then x0 is a fixed pointfixed point.
If xn = x0 for some n = 1, 2, ... we say that x0 is a periodic pointperiodic
point. If n is the smallest natural number with this property then n is said
to be the period period of x0 . We denote the corresponding cyclecycle by
γ = (x0 , x1 , ..., xn−1 ). In particular, the fixed point x0 is the periodic point
of period 1. Obviously x0 is a fixed point of the iterated map f n if x0 is a
periodic point of period n.
A fixed point x0 is called an attractorattractor if there exists a neighbor-
hood (ball) V (x0 ) of x0 such that all points y ∈ V (x0 ) are attracted by x0 , i.e.,
limn→∞ yn = x0 . If x0 is an attractor we consider its basin of attractionbasin
of attraction
A(x0 ) = {y ∈ Zp : yn → x0 , n → ∞}.
A cycle γ = (x0 , x1 , ..., xn−1 ) is said to be an attractor if x0 is attractor
of the map f n .
We have to be more careful in defining a p-adic analogue of a Siegel disk
(compare with the real or complex case [25]).
Let a ∈ Zp be a fixed point of a function f (x). The ball U1/pk (a), k =
0, ±1, ±2, ..., is said to be a Siegel disk (see [111]) if each sphere S1/pl (a),
where l ≥ k, is an invariant sphere of f (x), i.e., if one takes an initial point
on one of the spheres S1/pl (a), where l ≥ k, all iterated points will also be on
it: ρp (xn , a) = ρp (x0 , a) = const for all n. The union of all Siegel disks with
their center in a is said to be a maximal Siegel disk. Denote the maximal
Siegel disk by SI(a).
The simplest dynamical laws are given by monomial functions

ψs (x) = xs , s = 2, 3, ...

17
(each branch of the p-adic tree is multiplied by itself s times, producing a
new branch).
Our investigations, see chapter 3 and [111], show that the dynamical
model based on a p-adic space can already have a rich structure for the func-
tions ψs (x) = xs , s = 2, 3, ..., on Zp . There exist attractors, Siegel disks,
cycles. A prime number p, the basis of the coding system, plays the role of
a parameter of dynamical system. Behaviour of iterations depends on this
parameter. In fact, by changing p we can crucially change behaviour: at-
tractors may become centers of Siegel disks and vice versa, cycles of different
length may appear and disappear... We remark that behaviour of a ‘2-adic
cognitive system’ can strongly differ from the behaviour of, for example, a
‘5-adic cognitive system’, despite the same dynamical function (for example,
x → x2 )..
There exist I-states, attractors xa , which attract whole balls of I-states
(i.e., groups of I-states which have a sufficiently long common root with
xa ). There we obtain the same solution starting with any I-state x0 which
is sufficiently similar to xa . It should be noted that a common element of
I-states x0 and xa may be practically invisible, because of incredibly long
lengths of the cording sequences for x0 and xa .
There also exist cyclic I-states. Starting from such an I-state we shall
obligatorily return to it after a number of iterations (of course, this number
may be very large).
There are also Siegel disks. These are spheres of I-states Sr (a) which
cannot be changed in the process of thinking. If a thinking dynamical system
has arrived at Sr (a) it will never leave Sr (a). In this case the consciousness
must stop the dynamical system in the unconscious domain and change an
initial condition or a function ψs (x) by changing the parameter s. On the
other hand, the structure of an information sphere Sr (a) is quite rich. It is
possible that a dynamical system (or systems) in the unconscious domain
works on the information sphere Sr (a) during the whole life, i.e., a I-space
is reduced to Sr (a). In this case we shall never have any new I-state which
does not belong to this configuration space.
There are also fuzzy cyclesfuzzy cycles of I-states, cycles of balls,

Ur1 (a1 ) → Ur2 (a2 ) → · · · → Urk (ak ) → Ur1 (a1 ).

If radii rj = p−n , then starting with an I-state x1 which has a common root
of length n with the I-state a1 , we obtain an I-state x2 which has a common
root of length n with the I-state a2 , ...; after k steps we again obtain an I-state
which has a common root of length n with the I-state a1 . Such fuzzy cycles
of I-states correspond to cycles in dynamics of associations (see chapter 2).

18
 Our mathematical model might explain some aspects of human psychol-
ogy. Social types are determined by the laws ψs (x) which describe the work of
dynamical systems and by the ability of the consciousness to control the work
of the unconsciousness (to change parameters s in the process of thinking).
We now start mathematical considerations. The main aim of these con-
siderations is to show that even the simplest dynamical systems on p-adic
trees can demonstrate extremely complex behaviour.
To find fixed points of functions ψn (x) = xn in Zp we have to solve the
equation
xn = x in Zp .
There is the trivial solution x = 0. Other solutions are solutions of the
equation
xn−1 = 1 in Zp .
We have to know which kth roots of unity belong to Zp (in fact, to the unit
sphere S1 (0)).
Let n and m be natural numbers. We denote the greatest common divisor
of these numbers by the symbol (n, m).
We shall use the following well known result [40]:
Proposition 4.1. The equation xk = 1 has g = (k, p − 1) different roots
in Zp .
Denote the kth roots of unity by
θj,k , j = 1, ..., g = (k, p − 1) (with θ1,k = 1).
We study behaviour the of the dynamical systems ψn (x) = xn , n = 2, 3, ..,
in Zp . It is evident that the point x0 = 0 is the attractor for these dynamical
systems. The corresponding basin of attractions A(x0 ) = U1− (0) ≡ U1/p (0) =
{x ∈ Zp : |x|p ≤ 1/p}. We remark that (as the p-adic valuation takes only
discrete values pk , k = 0, ±1, ±2, ...) the ball U1 (0) is the union of the ball
U1/p (0) and the sphere S1 (0). Thus we need only to study the behaviour of
this dynamical system on the unit sphere S1 (0) = {x ∈ Zp : |x|p = 1}.
Theorem 4.1. The dynamical system ψn (x) = xn has m = (n − 1, p − 1)
fixed points aj = θj,k , j = 1, ..., m, on the sphere S1 (0). Fixed points aj 6= 1
belong to the sphere S1 (1).
1. If (n, p) = 1 then all these points are centers of Siegel disks and
maximal Siegel disks SI(aj ) are equal to the balls U1/p (aj ). For any k = 2, 3, ...
all k-cycles are also centers of Siegel disks of radius 1/p.
2. If (n, p) 6= 1 then all these points are attractors and A(aj ) = U1/p (aj ).
For any k = 2, 3, ... all k-cycles are also attractors.
The proof of this theorem can be found in chapter 3.

19
5 Role of Hierarchical Structure
Our mathematical model is based on two cornerstones which are (more or
less) independent.
The first is the assumption that the coding system which is used by
the brain for recording vectors of information generates a hierarchical struc-
turehierarchical structure between digits of these vectors. Thus if

x = (α1 , α2 , ..., αn , ...), αj = 0, 1, ..., m − 1,

is an information vector which presents in the brain an I-state, then the
digits αj have different weights. The digit α0 is the most important, α1
dominates α2 , ..., αn , ..., and so on.
The second is the assumption that the functioning of the brain is not
based on the rule of reasonrule of reason. The unconsciousness contains dy-
namical systems fs (x) which produce new I-states practically automatically.
This dynamics is controlled by a mechanical change of the parameter s and
the initial condition x0 . This control is provided by the consciousness.
The hierarchical structure in vectors of information implies a nearness
between I-states which corresponds to a length of the common root of these
I-states. This hierarchical structure induces the m-adic metric on the space
of I-states. Hence the description of I-states by m-adic numbers is a con-
sequence of the hierarchical structure on the space of I-strings. We may
consider other mathematical models of the process of thinking which will
be based on spaces of I-states without any hierarchical structure, i.e., use
other metrics on spaces of information. Here we can use other theories of
dynamical systems.
The standard metric on the space of I-strings x = (α0 , α1 , ..., αN ), αj =
0, 1, is the Hamming metric This metric is widely used in the theory of neural
networks and corresponding models of neuron activity, see, for example, [11],
[12].. It is defined as
N
X
ρ(x, y) = |αj − βj |, x = (αj ), y = (βj ).
j=0

However, this metric is not connected with a hierarchical structure. This

is not an ultrametric. The corresponding geometry strongly differs from the
geometry of the 2-adic tree. It might be that the Hamming metric is useful for
the description of physical dynamics in the brain (the propagation of electric
signals). The m-adic and more general hierarchical metrics are useful for the
description of cognitive dynamics in the brain (flows of associations, ideas,
desires, and emotions).

20
 We now discuss the role of the parameter s in a dynamical law fs (x).
The structure of a set S of parameters s is quite complicated. The value of
s is determined by physiological and psychological states of our brain and
an informative state of the brain (i.e., information which has been collected
by the brain up to the moment of the beginning of a thinking process) s =
(sphys , spsych , sinf ) ∈ S = Sphys × Spsych × Sinf . The physiological state sphys
depends on a large number of different conditions: food, weather, level of
our activity,... . All these conditions have the physical matter. From the
mathematical point of view this sequence of conditions can be described (at
least in principle) by an m-adic number in which each digit describes the
numerical value of the corresponding condition, sphys = β0 + β1 m + · · · +
βn mn + · · · . In the simplest case of the ‘yes’-’no’ coding, i.e., m = 2, we
may use the following description of sphys : β0 = 0 – ‘not hungry’, β0 = 1 –
‘hungry’; β1 = 0 – ‘not want to sleep’, β1 = 1 – ‘want to sleep’, and so on.
Of course, such a description strongly depends on an individual. For other
individual it may be more important to sleep than to eat. In this case we use
the digit β0 for the description of the sleeping state and the digit β1 for the
description of the nutritive state. The description of the psychological state
spsych is more complicated problem. We might describe spsych as psychological
states: ecstasy, depression,... . However, this classification is too rough. It
seems to be that there is practically an infinite number of psychological states.
It is important that the parameters sphys , spsych and sinf are not indepen-
dent: spsych depends on sphys and vice versa, both these parameters depend
on sinf , and, moreover, sinf may depend on spsych and sphys . Thus there is a
constraint
F (t, sphys , spsych , sinf ) = 0, (9)
which connects these parameters. Here t is the time parameter. The function
F strongly depends on an individual.
We now consider examples which illustrate some aspects of human psy-
chology.
Example 5.1 (Depression) Depression is characterized as a condition
of helplessness and ineffectiveness in which much of a person’s instrumental
behaviour extinguishes. We try to provide a mathematical description of
such a psychological state. Let the I-space of an individual I have a quite
complicated structure; for example, X = Zp , where p is large. Let us assume
that the dynamical system of the unconsciousness of I has the form ψs (x) =
xs . Then there are m = (s − 1, p − 1) fixed points a1 , ..., am on the sphere
S1 (0) and the fixed point a0 = 0. We interpret a1 , ..., am as ‘living stimuli’ of
I and the point a0 = 0 as the ‘nonliving stimulus’.
Let us consider the case (s, p) 6= 1. So the value s of the parameter

21
determining the dynamical law is divisible by the value p of the parameter
determining the coding system. Here all points a1 , ..., am are attractors.
Starting with any initial condition x0 ∈ S1 (0) I will arrive at one of
the living stimuli a1 , ..., am . This behaviour corresponds to a non-depressed
individual. The positive experience of the use of the initial conditions x0 ∈
S1 (0) implies that I prefers to use these initial conditions.
Now let us consider the case (s, p) = 1. In this case all points a1 , ..., am
are centers of Siegel disks. Starting with any point x0 ∈ S1 (0) I will be in-
volved in a practically infinitely long walk around center of the corresponding
Siegel disk. After a number of iterations the consciousness changes an initial
condition. On the basis of the previous positive experience the consciousness
prefers to take a new initial condition on the sphere of living stimuli S1 (0).
Such a choice will imply again and again a chaotic walk around one of the
living stimuli. Psychologically this situation is nothing than ‘helplessness’
and ‘ineffectiveness’. After a number of such negative experiments the con-
sciousness begins to use only initial conditions x0 belonging to the basin of
attraction A(a0 ) = U1/p (0) of the nonliving stimulus. This is nothing other
than the extinction of instrumental behaviour.
Our mathematical model of depression can be connected with well known
physiological models of depression (see, for example, [177], [1], [164], [160],
[173]). Physiological models correlate depression with a relative deficiency of
the catecholamine transmitter substances at synapses in the limbic system.
It seems that depression is connected with a deficiency of noradrenaline in
limbic sites. Let us introduce a parameter sna = 0, 1, where the values 1 and
0 correspond, respectively, to a normal level and deficiency of noradrenaline
(sna ∈ Sphys ). We now assume that the dynamical system of the human
unconsciousness is described as

fsna (x) = xp+sna −1 . (10)

The value sna = 1 gives the dynamical law ψp (x) = xp , which implies the
normal behaviour. On the contrary, the value sna = 0 gives the dynamical
law ψp−1 (x) = xp−1 which implies depression. This is the p-adic dynamical
description of the standard physiological model of depression. However, on
the basis of our p-adic dynamical theory we can essentially extend the model.
It is well known that the treatment of depression by the use of drugs which
change levels of corresponding physiological parameters (in particular, sna ) is
not successful in all cases. There is a quite high percentage of those ill whose
depressive behaviour cannot be changed by such a treatment. We explain
these results of the treatment by the following mathematical model. As we
have discussed, the set of parameters s which determine a dynamical system

22
of the unconsciousness contains not only physiological parameters sphys but
also psychological and informative parameters, spsych and sinf . In our model
of depression sna is a physiological parameter. Let us consider a psychological
parameter sdep = 0, 1 which is equal 0 and 1, respectively, in depressive and
non-depressive states, respectivelyThis psychological parameter can coincide
with one of the I-parameters. For example, let I be unemployed. Here
sdep = 0 if there is no positive information about a job and sdep = 1 if there
is some positive information about a job.. We assume that the dynamical
law has the form:
fsna ,sdep (x) = xp+sna +sdep −2 . (11)
If sdep = 0 then both values of sna imply depressive states which are de-
scribed, respectively, by laws ψp−2 (x) = xp−2 and ψp−1 (x) = xp−1 . If sdep = 1
then the value sna = 0 still implies depression, but the value sna = 1 implies
the normal active state.
Example 5.2. (Learning) In our model the process of learning is de-
scribed in the following way. Let us consider the following experiment. An
animal or human starts with the initial condition x0 and finds (in the pro-
cess of functioning of the dynamical system fα (x)) the answer xa which im-
plies a reward. Here xa is an attractor for the dynamical system fα (x) and
x0 ∈ A(xa ). For example, let us consider the standard experiment with rats:
food-deprived rats rewarded for pressing a bar by being presented with a
food pellet. Here
x0 =(to find food) and xa =(to press a bar)
(it is easy to code these problems in the 2-adic system). After a num-
ber of such ‘stimulus-response’ experiments animals and humans begin to
find the solution xa more quickly. This is learning. How can we explain
this phenomenon? In the process of learning, i.e., numerous repetitions of
a ‘stimulus-response’ experiment, the consciousness finds a new value α0 of
the parameter s (which determines a dynamical system of the unconscious-
ness) such that iterations of the dynamical system fα0 (x) which start with x0
converge to xa more quickly than the corresponding iterations of fα (x). Of
course, it is assumed that xa is an attractor of the dynamical system fα0 (x)
and x0 ∈ A(xa ). Moreover, the consciousness creates a functional dependence
α0 = g(x0 ) between the initial condition x0 which was used as the basis of
learning and the parameter s which determines the dynamical law.
For example, let p = 2, ψs (x) = xs , x0 = 10 = (01010...), α = 2. Here
x0 ∈ A(xa ) with xa = 0. Learning is a series of experiments in which starting
with x0 = 10 the dynamical system ψ2 (x) = x2 of the unconsciousness finds
the solution xa = 0. In this process the consciousness, for example, makes
the change α → α0 = x0 . Thus in this case g(x0 ) = x0 . Iterations of the

23
dynamical system ψg(x0 ) (x) = x10 converge to xa faster than iterations of the
original dynamical system.
Example 5.3. (Habituation) An ‘orienting response’ is defined as any
modality-non-specific response to novel stimuli which disappears after re-
peated presentations of the stimulus. The phenomenon of disappearance of
an orienting response is called ‘habituation’. We present the following exam-
ple to illustrate this phenomenon. We assume that the space of I-states has
the structure of Zp and, a dynamical law which gives a response to a novel
stimulus has the form ψs (x) = xs , where s is sufficiently large; the response
is the attractor xa = 0 and the stimulus x0 ∈ U1/p (0). Further, let repeated
presentation of the stimulus transfers this dynamical law into the dynamical
law of the form ψs−n (x) = xs−n where n is a number of repetitions of presen-
tations of the stimulus. It is evident that repetitions imply the use of a less
and less powerful dynamical system. In particular, iterations of the stimulus
x0 converge to the response xa slower and slower. In our model when n will
approach the value s − 1 the dynamical law will be ψ1 (x) = x. There the
convergence of iterations will disappear.

6 Role of Chance in Processing of Cognitive

Information
Let us discuss the role of chance in the process of thinking. To do this we
need to use the theory of random dynamical systems (see, for example, [15]
for general references about such dynamical systems). Often when starting
with the same initial condition we can obtain different results or no result at
all. This phenomenon is a consequence of a random behaviour of parameters
s. Random fluctuations of these parameters imply fluctuations in the form of
a dynamical law. In fact, s = s(ω) is a random variable and the functioning
of the unconsciousness is described by the random dynamical system

xn+1 (ω) = fs(ω) (xn (ω)), (12)

where the initial condition can also be a random variable, x0 = x0 (ω), see
chapter 4 for general random cognitive dynamics. The origin of this random-
ness is quite clear for physiological parameters. This is the standard random
behaviour of physical characteristics of a complex system. As a consequence
of the constraint (9) this physiological randomness implies a random be-
haviour of psychological and I-parameters. However, these parameters may
have their independent fluctuations. In general there is a random analogue

24
of constraint (9),
F (t, ω, sphys (ω), spsych (ω), sinf (ω)) = 0. (13)

Example 6.1. (Homeostatic states) It is well known that animals and

humans maintain themselves in a state of energy balance. That is, the net
amount of energy they absorb from radiant and thermal energy and from
the chemical energy of their food is equal to the amount of energy they
expend in their activities. In fact, not only is energy balance maintained,
but are a variety of other homeostatic conditions; for example, a constant
level of vitamins and minerals. How can we explain these balances by our
mathematical model? A homeostatic state is nothing other than an attractor
G of a dynamical system f bal (x) which controls the corresponding balance.
There is an extended basin of attraction A(G). Any choice of x0 ∈ A(G)
implies that after a number of iterations by f bal (x) the homeostatic state xa
will be recovered. Moreover, in our mathematical model we can assume that
x0 = x0 (ω) is a random variable (distributed over A(G)) which describes
fluctuations of quantities involved in the corresponding balance.
Of course, the regime of continuous functioning of all dynamical systems
bal
f (x) (corresponding to different homeostatic states) would be not useful for
the organism. We can interpret the well known physiological investigations
based on feedback models (originated in [18]; see also, for example, [46]) in
the following way. A dynamical system f bal (x) (corresponding to G) begins
to work if a deviation from G is sufficiently large.
We now consider a homeostatic state G controlled by the dynamical
system f bal (x). There exists a threshold  > 0 such that if the distance
ρ(x0 , G) >  then the dynamical system f bal (x) begins to work starting with
x0 . If A(G) = Ur (G) then, of course,  ≤ r. In fact, we can explain destruc-
tion of a homeostatic state G by the change of the threshold  :  → 0 > r.
The same model can be applied to describe more complicated balances. In
principle there may be two or more stable states. For example, human sexual
activity has three stable states Gh , Gn , Ga which correspond to high, normal
and zero levels of this activity. All these states are stable and have quite
extended basins of attraction. Therefore it is complicated to leave one of this
states and move to other.
Example 6.2 (Manic depressive behaviour) By using stochastic dynam-
ical systems we might provide a mathematical description of fluctuations
between the states of depression and high excitation and activity. Let us
assume that the parameter sna which has been introduced in Example 5.1
is a random variable, sna = sna (ω), with a probability distribution P(ω :
sna (ω) = α) = qα , α = 0, 1. If the dynamical law of an individual I has

25
the form fsna (ω) = xp+sna (ω)−1 then I oscillates between the state of depres-
sion (sna (ω) = 0) and the normal state (sna (ω) = 1). Let the behaviour
n
of I be controlled by the dynamical system gsna (ω) = xp +sna (ω)−1 , where n
is sufficiently large. Then the state sna (ω) = 1 is characterized by an in-
credible activity and excitation of I. The probability distribution of sna (ω)
determines shares of depression and excited periods in the life of I. For ex-
ample, if q0 = 9/10 and q1 = 1/10 then I is practically always depressed; if
q0 = 1/10 and q1 = 9/10 then I is a practically normal personThe parameter
n also plays the important role in the creation of psychological behaviour
of τ. If the number n is very large and the probability q1 ≈ 1, then the
τ will be permanently in the state of the abnormal excitation.. Of course,
the dynamical law may depend not only on physiological parameters such
as sna (ω), but also on psychological parameters which are also assumed to
be random variables. Here, for example, the behaviour of I can be de-
n
scribed by the law φsna (ω),sdep (ω) = xp +sna (ω)+sdep (ω)−2 , where, for example,
sdep (ω) = 0, 1 with probabilities c0 , c1 . To introduce the time factor into the
investigation of manic depressive behaviour we can use stochastic processes,
sna (t, ω), sdep (t, ω), instead of random variables. It would be natural to use
the Poissonian process and other jump processes.

7 Information Reductionism
We have presented the mathematical model of the process of thinking. As
we have seen, this model might explain some psychological phenomena. May
we use this model for some practical purposes? It seems that our theoretical
considerations might be developed to a practical level.
Let us assume that we have found a dynamical law fs (x) which describes
the functioning of some thinking processor in the unconscious domain of an
individual I. Then, in fact, we can predict (and control) a behaviour of I.
For example, we propose for I an initial idea x0 belonging to the basin of
attraction A(xa ). Then we can be sure that I will obtain xa after some period
of thinking. Our mathematical model gives the rule for finding x0 : this idea
(the I-state in the present simplest cognitive model) must have a common
initial segment with xa in the coding representation. Probably this method
might be used in psychiatry. The behaviour of a patient can be corrected
by giving him initial ideas x0 which are quite attractive to him but (as we
know on the basis of his dynamical law fs (x)) will obligatorily produce the
solution xa which can be considered as a treatment of mental illness.
Of course, the main problem is to find the right dynamical law fs (x).
We remark that fs (x) is nothing other than a mathematical description of

26
the functioning of the unconsciousness. Thus we cannot say that the un-
consciousness use the law fs (x). The law fs (x) describes only (with some
approximation) some aspects of this functioning. How can we find this ap-
proximative law? The only possibility is to find reactions (results of the
functioning) for a large class of initial conditions x0 . Then we compare these
reactions with the behaviour of dynamical systems which were investigated
by theoretical considerations. If, for example, an essential part of these re-
actions coincide with the corresponding results for the dynamical system
x → x2 then we may use this law to describe a behaviour of this individual
I. In the ideal case it should be possible to realize the following experiment.
There are two rooms: I is in one room and a computer with a program for
the dynamical law fs (x) of I is in another room. An initial I-state x0 is
given to I and the computer. Then (after some time, which may be different
for I and a computer) we can find the same answer y in both rooms.
The problem of finding the form of fs (x) for I is very complicated because
of the large spectrum of the parameters s which determine this law. However,
it seems to be possible to find this form for some values of physiological,
psychological and other I-parameters.
In fact, our investigation is a kind of reductionist psychological theory
(compare, for example, with [176], [133], [54], [46]). Reductionist theories
can have different cornerstones. For example, it can be a reduction to ag-
gressive instincts and the ethological principles governing them, or to sup-
pose innate behavioral tendencies to perpetuate one’s genotype, or to give
neurophysiological accounts of behaviour (see [176], [133], [54], [46]). We
reduce psychological and social behaviours to properties of the coding and
dynamical systems of brain. Investigations devoted to the neurophysiologi-
cal reduction (see, for example, [46]) are closely connected with our model
If we imagine the brain as a kind of computer then the neurophysiological
model describes the hardwarehardware of this computer and our mathemat-
ical model describes the softwaresoftware of this computer (or at least the
essential part of this software).. Adherents of neurophysiological models try
to explain psychological phenomena as the results of different states for the
physiological parameters sphys . We accept that these parameters play an im-
portant role in psychological phenomena. We use these parameters in our
model to determine the forms of dynamical systems of the unconsciousness.
However, we think that neurophysiological models have to be supported by
the corresponding mathematical apparatusIn the same way we cannot explain
the results of a computer functioning only by a description of the physical
processes in its hardware..
Our model can be applied to other living organisms. Their thinking
processes can also be described as m-adic dynamical systems. The main

27
difference between the thinking processes of people and other organisms is
the stronger separation between the consciousness and unconsciousness in
the human brain. The most primitive living organisms have a small volume
of consciousness. Iterations of a thinking dynamical system are not hidden.
Each step xn is considered as the order with its immediate realization. For
example, such a kind of behaviour is demonstrated by insects insect which
carry out step by step all orders xn of their dynamical systems. The small
volume of consciousness does not give the possibility of switching a regime
of the dynamical system (to change a parameter s of fs (x)). Therefore,
for example, insects may repeat the same behaviour many times if they start
with the same initial condition. They do not have additional cognitive blocks
to analyse results of the process of thinking.
Further more, we apply our model to explain similarities in a particular
psychological behaviour of different species. These similarities can not be ex-
plained by physiological arguments, because the structures, for example, in
a fish’s brain, fulfilling particular psychological functions are different from
those doing the job in a rat’s brain [46]. However, these similarities can
be explained by using dynamical systems in the space of I-states. As we
have demonstrated in the previous examples, a psychological state ψ can be
described by the functioning of the corresponding dynamical system fsψ (x).
Therefore it is not important what particular neural structure realizes func-
tioning of fsψ (x). The only important thing is the form of the mathematical
function fsψ (x). In the same way it is not important what kind of computer
we use to realize the simulation of the functioning of fsψ (x). If the mathe-
matical law (for example, fs (x)) is fixed an ‘IBM-PC’ and a ‘Macintosh’ will
produce the same result. If the brains of a fish and a rat realize the same
mathematical law fsψ (x) then fish and rats demonstrate the psychological
state ψ.
Furthermore, our model might be useful to find a solution of the following
problem of physiological psychology. There are numerous evidences that
often it is not easy to ‘localize psychological function’ in the neural system
[142], [84], [137], [82], [48], [46], [52], [53], [81], [146]. For example, if one
tries to specify the particular structures involved in unlearning instrumental
conditioning, one finds that in rodents this function is subserved by the
septal-hippocampal system, whilst in primates the orbito-frontal cortex takes
on that role [84]. Moreover, the same state ψ can be a result of the functioning
of different neural structures at different instants of time, see [142], [84],
[137], [82], [48], [46], [52], [53], [81], [146]. This situation is explained by the
evident fact that the same dynamical law fsψ (x) can be realized by different
neural structures at different instants of time. The same model might explain
similarities in social behaviour of different species.

28
 Our m-adic model of the process of thinking have (at least indirect) rela-
tions to investigations which were performed in neuro and psycho physiology
– see, for example, [9], [10], [19], [23], [41], [42], [48], [51]-[55], [65], [71],
[81], [82], [98], [103], [104], [138], [144], [146], [149], [153], [154], [158], [159],
[170],[171], [174], [179], [180], [183], [193], [198]. Those investigations support
(at least partially) following ideas used in our model:

1) The split of I-processing in brain into unconscious and conscious pro-

cessing, [78], [23], [41], [42], [51], [65], [69], [76], [81], [138], [146],[153], [158],
[174], [180], [183] ;
2) Association based thinking, [158], [96], [146].
[Hierarchy of Information] Hierarchy of Information
The idea that some essential features of the process of thinking can be
simulated by neural networks is quite popular in the cognitive community;
see, for example, [11], [12], [93]. As with all theories, such an approach
has strong as well as weak aspects. In particular, cognitive models based
on neural networks can be criticized from the positions of artificial intelli-
gence [185] (based on the computational, Turing’s machine [184], viewpoint of
thought) or quantum cognitive theories, [148], [100], [150], [175] and chapters
5,6 (based on a field-like representation of thought), or cognitive architecture
(based on the necessity of a linguistic medium, [77], [155], [199]), see [185]
on the extended discussion. However, it seems that, despite all the critique,
neural networks are very useful for describing (or at least simulating) the
process of thinking.
Regarding the critique from positions in artificial intelligence,artificial in-
telligence it seems that the role of Turing-like computations in the process of
thinking is overestimated (compare to [150]). The brain may contain some
thinking blocks based on transformations of information performed by Tur-
ing machinesTuring machine (recursive functions)recursive functions. How-
ever, there is no evidence that the brain contains only Turing (algorithmic)
thinking blocks. It is easy to present mathematical models which show that
numerous cognitive features can be demonstrated by dynamical networks
performed by non-Turing transformations, chapter 1.
Regarding quantum cognitive models, I agree that the process of thinking
could not be reduced to the functioning of neural networks. Therefore neural
network blocks of a brain must be considered as only a part of the thinking
machine. This part of brain (which we called subconsciousness in [113])
is responsible for the unconscious flow of information in brain. In some
sense this is a realization of generalized artificially intelligence postulate:
a part of the brain is a computer-like machine, but this machine uses not
only recursive, but also non-recursive transformations of information. In

29
chapter 1 we have demonstrated that natural classes of such non-recursive
transformations (continuous functions) can be induced by the hierarchic m-
adic topology on the information space of a cognitive system τ .
The brain has to have another information structure (which we call con-
sciousnessconsciousness) which is responsible for analysis, ‘understanding’,
of results of the functioning of unconsciousness. Information becomes con-
scious only at the higher levels of a thinking machine. A mechanism of the
functioning of the consciousness is still the greatest mystery (see, for example,
Descartes [59], G. Ryle [161], D. Bennett [58], D. Chalmers [44] for numerous
discussions on various aspects of consciousness).
Regarding the linguistic critique of the neural networks approach, I agree
that this approach must be essentially improved to take into account the
semantic and syntactic structures of thought, specially:
(a) productivity - the capacity to produce a huge set of utterances from
a finite set of information strings;
(b) inferential coherence - the ability to infer, for example, from the fact
that ‘Mary and John went to the shop’ that ‘Mary went to the shop’ and
that ‘John went to the shop’ (see [77], [185] for details).
In chapter 1 we studied cognitive models, in that unconscious processes
were (mathematically) represented by continuous (with respect to a tree
topology) maps f : XI → XI . In deterministic models a new information
state xn+1 ∈ XI was obtained on the basis of the previous state xn ∈
XI : xn+1 = f (xn ). In random dynamical models the game of chance
(with a chance parameter ω) changes one thinking block to another: xn+1 =
fzn (ω) (xn ), where { fz1 , . . . , fzK } is a set of ‘thinking functions’ and zn (ω)
describes the game of chance. In the theory of random dynamical systems
zn (ω) = g(θn−1 ω), where θ is a flow of chance. One of the main conse-
quences of mathematical investigations [68] is that the flow of chance θ plays
an extremely important role in the proceeding of information. It seems that
ω could not be just reduced to physical and informational noise. The ω is
merely a (statistical) regulator of the choice of different thinking blocks in
the unconsciousness.
In this chapter we improve this dynamical cognitive model on hierarchic
information trees. We try to reply to the linguistic critique by introducing a
new hierarchy:
information states x → associations A → ideas J.
Here x belongs to a tree XI ; an association A is an equivalence class of
information states; an idea J is a set of associationsWe remark that the basic
space of I-states has its own hierarchy, namely, the hierarchy of an m-adic
tree.. For some class of maps f : XI → XI , the dynamics of I-states induces
dynamics in the spaces of associations and ideas, XA and XID , respectively

30
(can be lifted to these spaces). Such a model with two axes of hierarchy
satisfies (more or less) the postulates (a) and (b) of the linguistic approach
to thought:
(a) a rather small set of information states can induce (via the forming
of associations) a huge set of inferences;
(b) each subset Je of an idea J is also an idea and by using the simplest
inference law: J ⇒ Je if Je ⊂ J, the consciousness can obtain a huge set of
inferences from each idea J.
A new property of the dynamics of ideas is that (for a large class of dy-
namical systems on m-adic trees) for each initial idea J0 its iterations are
attracted by some idea Jattr . The latter idea is considered by the conscious-
ness as a solution of the problem J0 . In the opposite to such an attractor-like
dynamics of ideas, the dynamics of information states (on the tree XI ) or
associations need not be attractive. In particular, there can exist numerous
cycles or ergodic evolution. In general the process of thinking at the level
of I-states (or associations) does not give a definite answer to each initial
problem x0 ( or A0 ). Starting with some initial data dynamical processors
on m-adic trees can produce cycles. In fact, even the simplest dynamical
systems (monomial maps) have a huge number of cycles of various lengths.
The distribution of these cycles is extremely sensitive to the basis m of a cog-
nitive coding system as well as to other parameters of a dynamical system,
see chapter 3 and [111], [119], [122], [145]. Hence a cognitive system τ which
tries to proceed thinking only at the primary mental level will be definitely
involved in chaos of numerous cycles without any possibility of finding the
definite solution. This cyclic chaos which is manifested on the ground and
first floors of the ‘thinking house’, the floors populated by I-states and asso-
ciations, disappears on the second floor: starting with any initial idea J0 ; a
dynamical processor always produces an idea-attractor Jattr .
Moreover, cyclic chaos on the cognitive ground floors (which are popu-
lated by I-states and associations) has a positive influence on the process
of thinking on the second floor (which is populated by ideas). Our math-
ematical study demonstrates that such a chaos of cycles is the basis of the
creativeness of cognitive systems. The richness of the set of ideas-attractors
(ideas-solutions) is based on the huge number of cycles at the level of I-states
and associations. Therefore:
Stronger chaos of cycles for I-states → stronger creativeness.
By using higher cognitive levels of the representation of information a
cognitive system strongly improves the regularity of thinking dynamics. The
random dynamics of ideas exhibits the same feature. The latter dynamics
exhibits a new property, namely, a long range (statistical) memory. In general

31
a new idea Jn+1 is correlated with all the ideas, J1 , . . . , Jn , produced in
previous iterations.
Finally, we note that the use of a new cognitive hierarchy (in combination
with the basic hierarchy of the m-adic tree) strongly improves the information
power of a cognitive system. We demonstrate by an example that a prim-
itive cognitive system which (at each instant of time) operates at the level
of I-states with 300 bits of information can operate with N = 4000, 000, 000
ideas. It may be that such a hierarchical information capacity gives the possi-
bility to cognitive systems on solving the problem of combinatorial explosion
(which is well known in cognitive sciences, see Dennett [58] for an extended
discussion of this problem in the creation of a ‘brain in a vat’). Moreover, it
might be that some cognitive systems could create ‘mental towers’ with an
infinite number of floors: I-states, associations of these states, associations
of associations (which we call ideas), associations of ... of associations,....
Another (rather unexpected) feature of the dynamics of ideas is the vio-
lation of the materialistic axiommaterialistic axiom of cognitive science, see
Wittgenstein [199]. In principle, the same idea J, a ‘state of mind’, can be
represented by very different configurations of excited neurons, ‘states of the
brain’.
We discuss a model of the hardware realization of hierarchical information
state x ∈ XI by considering so called hierarchical chains of neurons. Such
a model can be based on the spatial domain model [11] as well as on the
frequency domain model [94]. In the first case hierarchical chains of neurons
are spatial chain-structures. In the second case a hierarchical chain of neurons
may have a dust-like spatial structure. It is a chain in the frequency domain.
Of course, our unconscious/conscious thinking model was motivated by
Freud [79]. It seems that he was right that not all information flows in
the brain are flows of conscious information. Even an idea-attractor Jattr
obtained via a ‘dynamical thinking’ is still unconscious. To become conscious
it must be analysed and in some sense ‘recognized’, by the consciousness.

8 Hierarchical Coding of Information

As every where in this book, the symbol τ will be used to denote a cognitive
system. Here we discuss a possible hardware realization of our m-adic I-trees
at the level of neuron nets. In our model it is assumed that each neuron n has
m > 1 levels of excitation, α = 0, 1, ..., m − 1 (for example, in the simplest
model α = 1/0, firing/not firing). In this model an individual neuron has no

32
cognitive meaning. Cognitive information is represented by chains of neurons,

N = (n0 , n1 , ..., nM −1 ).

Each chain of neurons N can (in principle) represent mM different I-states

(information strings):

x = (α0 , α1 , ..., αM −1 ), α ∈ {0, 1, ..., m − 1}, (14)

corresponding to different levels of excitement for neurons in N . Denote the

set of all possible I-states by the symbol XI .
In our model each chain of neurons N has the hierarchic structure:
neuron n0 is the most important,
neuron n1 is less important than neuron n0 ,...,
neuron nj is less important than neurons n0 , ..., nj−1 .
This hierarchy is based on the possibility of a neuron to ignite subsequent
neurons in this chain:
n0 can ignite all neurons n1 , ..., nk , ..., nM −1 ;
n1 can ignite all neurons n2 , ..., nk , ..., nM −1 ,
and so on; but the neuron nj cannot ignite any of the previous neurons
n0 , ..., nj−1 . Moreover, the process of igniting has the following structure. If
nj has the highest level of excitation, αj = m − 1, then increasing αj by one
unit induces the complete relaxation of the neuron nj , αj → αj0 = 0, and
increasing by one unit the level of excitation αj+1 of the next neuron in the
chain,
0
αj+1 → αj+1 = αj+1 + 1 . (15)
If the neuron nj+1 already was maximally excited, αj+1 = m − 1, then the
transformation (15) will automatically imply the change by one unit of the
state of neuron nj+2 (and the complete relaxation of the neuron nj+1 ) and
so onIn fact, the transformation (15) is the addition with respect to mod m..
We shall use the abbreviation HCN for hierarchical chain of neurons.hierarchical
chain of neurons
In principle the same I-state x can be represented by a few distinct HCN s
which have the same levels of exciting for corresponding neurons. Such a
performance can essentially increase the safety of I-processing. On the other
hand, it may be one of the possible sources of spatial non-locality of mental
functions. In principle, HCN s producing the same I-state x can be located
in spatially distinct domains of the brain.
Let N , M, L be HCN s producing I-states x, y, z respectively. Denote by
k(N , M) (k(N , L) and k(M, L)) length of an initial segment in chains N

33
and M (N and L, M and L) such that corresponding neurons in N and M
have the same levels of exciting. Then it is evident that

k(N , M) ≥ min[k(N , L), k(L, M)]. (16)

This gives the strong triangle inequality for the m-adic metric.
Special collections of HCN s form new cognitive objects, associations.associations
Let s ∈ {0, 1, ..., m − 1}. A set

As = {x = (α0 , ..., αM ) ∈ XI : α0 = s} ⊂ XI

is called an association of the order 1. This association is generated by a col-

lection Bsτ of all HCN s N = (n0 , n1 , ..., nM −1 ) which have the state α0 = s
for neuron n0 . If the set Bsτ is empty the association As does not exist in
the brain (at this instance of time). Regarding the problem of localization of
mental functions in brain it must be noted that chains of neurons participat-
ing in the creation of the same association can be located in distinct domains
of brain. In some sense there is double nonlocalty: nonlocality on the level
of I-states (which is produced by spatially extended chains of neurons) plus
nonlocality on the level of associations (which is produced by participation
of different chains of neurons in the creation of the same association).
Associations of higher orders are defined in the same way. Let s0 , ..., sl−1
belong to {0, 1, ..., m − 1}, l ≤ M. The set

As0 ...sl = {x = (α0 , ..., αM ) ∈ XI : α0 = s0 , ..., αl−1 = sl−1 }

is called an association of order l. Such an association is represented by a set

Bsτ0 ...sl ⊂ B τ of HCN. We remark that associations of the order M coincide
with I-states for an HCN. We shall demonstrate that a cognitive system τ
obtains large advantages by working with associations of orders l << M.
Denote the set of all associations of order l by the symbol XA,l . We set

XA = ∪l XA,l .

This is the set of all possible associations. We remark that the space of
I-states XI is embedded in the space of associations XA .
Sets of associations J ⊂ XA also have to have a cognitive meaning. For
example, let A = Aα0 ...αl be an association of the order l. This is a set
of I-states having the special structure. Let us consider the set-theoretical
complement of A in the I-space XI : A = {x ∈ XI : x 6∈ A}. Such a set of I-
states has an evident cognitive interpretation: it is negation of the association
A. For example, if A is an association on a sunny day, then A is the set of
all images which are not related to an image of a sunny day. We mention a

34
simple mathematical fact: the set A can also be represented as a family of
associations of the order l. These sets, associations, have empty intersections.
Hence such families of associations also must have some cognitive meaning.
We obtain the same conclusion by using the logical operation ‘or’. The logic
operation ‘or’ is realized in XI as the set-theoretical union. We remark that
the logical operation ‘and’ is trivial on the space of associations XA : if
A ∩ B 6= ∅, then A ∩ B = A or B.
Sets of associations will be called ideasideas (of order 1). Denote the set
of all ideas by the symbol XID . In principle, it is possible to consider sets
of ideas of order 1 as new cognitive objects (ideas of order 2) and so on.
However, we restrict our attention to the dynamics of ideas of order 1.
The space XID consists of points-associations. Roughly speaking at the
level of ideas we can forget about the internal I-structure of associations.
The space XID is endowed with the standard structure of a Boolean algebra:
‘or’- union, ‘and’-intersection, ‘or’- complement. Thus our cognitive model is
closely related to the calculus of propositions of Boole [38], [39]. The main
distinguishing feature of our model is the presence of the m-adic internal
structure for elements of the Boolean algebra.
It is natural to assume that the representation of cognitive information
on the level of ideas plays the crucial role in mental functions (at least such
‘high level functions’ as, for example, emotions or goal-directed behaviour).
This induces an additional spatial nonlocality of mental functions, since an
idea can be formed by distinct associations which are in general formed by
spatially separated chains of neurons (which are also extremely nonlocal ob-
jects).
Remark 1.1. (Associations, ideas and complexity of cognitive behaviour)
One of the main features of our model is that not only I-states, x ∈ XI , but
also associations, A ∈ XA , and ideas, J ∈ XID , have a cognitive meaning.
One of the reasons for using such a model is that complex cognitive behaviour
can be demonstrated not only by living organisms τ which are able to pro-
cess in ‘brains’ large amounts of ‘pure information’ (I-states), but also by
some living organisms with negligibly small ‘brains’. It is well known that
some primitive organisms τpr having (approximately) N = 300 nerve cells can
demonstrate quite complex cognitive behaviour: ability for learning, complex
sexual (even homosexual) behaviour. Suppose, for example, that the basis
m of the coding system of τpr is equal to 2. Here each nerve cell n can yield
two states: 0, non-firing, and 1, firing. Non-hierarchical coding of informa-
tion gives the possibility of performing in the brain (at each instant of time)
300 bits of information. In the process of ‘thinking’ τpr transforms these
300 bits into another 300 bits. It seems that such 300-bit I-dynamics could

35
not give a complex cognitive behaviour. We now suppose that τpr has the
ability to create hierarchical chains of nerve cells (a horizontal hierarchy).
Let, for example, such HCN s have length L = 5. Thus τpr has N = 60
HCN s (so the set B τpr has 60 elements). The total number of I-states,
x = (α0 , α1 , α2 , α3 , α4 ), αj = 0, 1, which can be performed by HCN s of the
length L = 5 is equal to NI = 25 = 32. Thus the brain’s hardware B τpr (the
collection of all hierarchic chains of neurons) can represent all I-states simul-
taneously (since NI < N ). Some I-states are represented by a few chains of
neurons (this increases the safety of the representation of information). We
assume that all I-states are performed by the brain at each instant of time.
We suppose that τpr is able to operate with associations and ideas. The τpr
have Na = 2k associations of order k = 1, 2, ..., 5. The number of homoge-
neous ideas (i.e., ideas that are formed by associations of the same order) of
τpr is equal
2 3 4 5
NID,hom = (22 − 1) + (22 − 1) + (22 − 1) + (22 − 1) + (22 − 1)
= 4295033103 >> 300
(each term contains -1, because empty sets of associations are not considered
as ideas). Hence τpr works with more than 4295033103 ‘ideas’ (having at the
same time only NI = 32 I-strings in his brain). Moreover, if we consider
the possibility of a cognitive interpretation of ideas of higher levels (sets of
ordinary ideas), then even primitive cognitive systems could operate with
fantastically large amounts of information. For example, in the case of τpr
the number of ideas of level 2:
(2)
NID > 24295033103 .
In our model the brain’s ‘hardware’ of τ is given by an ensemble B τ of
HCN s. For an HCN N ∈ B τ , we set i(N ) = x, where x is the I-state of N .
The map i : B τ → XI gives the correspondence between states of the brain
and states of the mind. In fact, the map i provides the connection between
the material and mental worlds. In general it may be that i(N1 ) = i(N2 ) for
N1 6= N2 . As we have already mentioned, this is one of the possible sources
of the spatial non-locality of mental functions. It is natural to assume that
in general the map i depends on the time parameter t : i = it . In particular,
if t is discrete we obtain a sequence of maps it : t = 0, 1, 2, .... Such a
time-dependence is also a source of spatial nonlocality of mental functions.
In the spatial domain modelspatial domain model each HCN N is a chain
of physical neurons which are connected by axons and dendrites, see, for ex-
ample, [70]. In principle, such a chain of neurons can be observed as a spa-
tial structure in the Euclidean space R3 . Unfortunately, the present level of

36
neurophysiological measurements is not sufficient to observe the functioning
(exiting) of chains of individual neurons in cognitive experiments.
In the frequency domain modelfrequency domain model [94] the digits
αj ∈ {0, 1, ..., m − 1} can be considered as the frequencies of oscillations
of the neurons nj , j = 0, 1, 2, ..., which form a ‘frequency HCN ’ N . Here
N need not be imagined as a connected spatial structure. It may have a
dust-like structure in R3 .
Remark 1.2. (Neural code), compare with [70]. In our model in-
formation is coded by long chains of neurons, HCN s. The dense struc-
ture of brain connections implies that different HCN s must have intersec-
tions. Thus the same neuron n can be contained in a few HCNs: Nj =
(nj0 , nj1 , . . . , njM −1 , . . .), j = 1, . . . , K, and njkj = n. Here the index kj is used
to describe the location of the same neuron n in different hierarchical chains.
The act of firing of the neuron n plays a different role in different HCN s,
N1 , . . . , NK . The firing of a single neuron has no direct cognitive meaning.
Therefore a neural code cannot be found by calculating frequencies of firing
for individual neurons. To find a neural code we must find HCN s.
Remark 1.3. (Mutual interference of thinking processors). Let us con-
sider two thinking processors π1 and π2 (for example, π1 is the sexual pro-
cessor and π2 is the religious processor). Suppose that both are one-chain
processors: πj works on single hierarchic chain of neurons
Nj = (nj0 , nj1 , . . . , njM −1 , . . .), j = 1, 2.
Suppose that these chains have nonempty intersection
(a, . . . , d) ∈ N1 ∩ N2 : a = n1ja = n2ia , . . . d = n1jd = n2id .
The frequencies of the firing of the neurons a, . . . , d are determined by
both processors, π1 and π2 . Such an intersection of HCN s corresponding
to different thinking blocks is the origin of the mutual influence of different
mental functions. Suppose that the dynamical function of the processor πj
has the form fπj (x) = xnj . Suppose that, for the sexual processor π1 , the
initial state x0 (induced, for example, by external information) is always
equal to 0 (no sexual input). Thus fπ1 (x0 ) = fπ1 (0) = 0 and the isolated
sexual processor π1 will always produce the zero output. However, if N1 ∩
N2 6= ∅, then the functioning of π2 may induce nonzero outputs of π1 . The
stabilization to an attractor in N2 implies the stabilization of the I-states of
all neurons in N2 . In particular, the neurons a, . . . , d will obtain fixed values
αa , . . . , αd . Thus π1 automatically produces a nonzero attractor

x = (0, . . . , αa , . . . , αd , 0, . . .).
Here a religious input gives a sexual output. Such an output will be rather
abnormal, because the digits αa , . . . , αd were produced via iterations of the

37
religious dynamical function fπ2 which can strongly differ from the sexual
dynamical function fπ1 .
Remark 1.4. (Conflict of dynamical systems). Let us consider again
the functioning of two dynamical systems which has been described above.
Suppose that the input or output channels of the processor π1 are blocked
for a long time. The neurons a = n2ia , . . . , d = n2id (on the intersection of the
sexual and religious hardware) do not receive any signal from neighbouring
neurons in N1 . At the same time these neurons a = n2ia , . . . , d = n2id send
signals (produced via the functioning of π2 ) to neighbouring neurons in N1 .
In principle, such a process can imply a reconstruction of thinking processors:
the chain of neurons N1 becomes a part of π2 . So π1 is conquered by π2 .
The mechanism of a ‘hardware’ realization of thinking maps f : XI → XI
has not yet been found on the basis of neurophysiology. We can propose
possible mechanisms. The simplest ‘hardware’ realization (on the basis of
HCN s) is the following: each HCN N having the I-state x ∈ XI changes
its I-state to y = f (x) ∈ XI . Here each HCN N performs (in the pro-
cess of thinking) a wide range of I-states. There is no spatial domain lo-
calization for states of mind. This ‘hardware’ is quite natural for a fre-
quency domain model. It may be that each HCN N could represent wide
range of combinations of frequency oscillations ν = (ν0 , ν1 , ν2 , ...), where
νj = (2παj )/m, α = 0, 1, ..., m − 1.
It may be that the ‘hardware’ basis for the maps f : XI → XI is more
complicated. Let each HCN N having an I-state x ∈ XI change the I-states
of a group
T (N , x) = (N1 , . . . , Nl )
of HCN s to a new I-state y = f (x). The main feature of such hardware
is that if i(N ) = i(M) = x, then i(T (N , x)) = i(T (M, x)) = y. Thus if
two HCN s, N and M, have (at some instant of time t = n) the same I-
state x ∈ XI , then both groups of HCNs, T (N , x) and T (M, x), (which are
ignited by N and M respectively) will obtain (at the next instant of time
t = n + 1) the same I-state y ∈ XI . We do not claim that a real brain has
such a property. Our aim is to perform some mathematical simulation on
the basis of such a model and see what features of cognitive behaviour can
be simulated.
A group T (N , x) of HCN s depends on a system of connections C between
neurons (hardware) as well as on an I-state x of HCN N .
Remark 1.5. (Materialistic axiom). If the consciousness of τ is insen-
sitive to the cardinality K(x) of the set i−1 (x) (the number of HCN repre-
senting the same state x of mind), then the dynamical evolution T (N , x, t),
K = K(t, x), violates the materialistic axiom. Here very different spatial

38
(and frequency) configuration of neurons can perform the same state of
mind. Such a model is not so unreal. It may be that the consciousness
takes into account only low bounds of K(t, x). Here all states of the brain
with K(t, x) ≥ Kmin have the same mental meaning. In general Kmin can
also evolve with time and depend on x: Kmin = Kmin (t, x).
In the mathematical model it is convenient to use infinitely long I-strings
instead of I-strings of a finite length M. We shall consider the I-space XI =
Zm and repeat all considerations of this section (in particular, definitions of
associations and ideas).

9 Flows of Associations, and Ideas

In this section we study the simplest dynamics of associations and ideas.
Such I-dynamics are induced by the corresponding dynamics of I-states, i.e.,
‘ruled’ by functions f : XI → XI which do not depend on time and random
fluctuations. This process of thinking has no memory: the previous I-state
x determines a new I-state y via the transformation y = f (x) :

xn+1 = f (xn ). (17)

Suppose that, for each association A, its image

B = f (A) = { y = f (x) : x ∈ A }
is again an association. Denote the class of all such maps f by the symbol
A(XI ). Sometimes we shall consider dynamics which is restricted to a subset
O of XI ; in such a case we shall use the symbol A(O).
If f ∈ A(XI ), then the dynamics (17) of I-states of τ induces the dynam-
ics of associations
An+1 = f (An ). (18)
Starting with an association A0 , τ obtains a sequence of associations A0 , A1 =
f (A0 ), . . . , An+1 = f (An ), . . .. We can say that the dynamics in the I-space
XI for transformations f ∈ A(XI ) can be lifted to the space of associations
XA .
The dynamics of associations (18) automatically induces the dynamics
of ideas J 0 = f (J) = {B τ = f (A) : A ∈ J}. Thus each idea evolves by
iterations:
Jn+1 = f (Jn ). (19)
Starting with an idea J0 , τ obtains a sequence of ideas: J0 , J1 = f (J0 ), . . . , Jn+1
= f (Jn ), . . .. The reader will see that there is a difference between the pos-
sibility of lifting to associations and to ideas: in the latter case it is always

39
possible. In fact, this difference is just a consequence of our definition of
ideas. Here we do not try to specialize classes of associations that form new
cognitive objects, ideas. In particular, we do not use any hierarchical struc-
ture in the forming of ideas on the basis of associations. In fact, we develop
such an approach only for one reason: to simplify the model.
Geometrically associations are represented as bundles of branches of the
m-adic tree. Ideas are represented as sets of bundles. So the dynamics (17),
(18), (19) are, respectively, the dynamics of branches, bundles, and sets of
bundles on the m-adic tree. The I-dynamics on Zm is generated by the
maps f : Zm → Zm . We are interested in maps which belong to the class
A(O), where O is some subset of Zm . We remark that there is a one to
one correspondence between associations of order l, As0 ...sl−1 , and balls Ur of
radius r = 1/pl in the metric space Zm , namely
As0 ...sl−1 = Ur (a), r = 1/pl , a = s0 + ... + sl−1 ml−1 .
Thus f ∈ A(O) maps a ball onto a ball: f (Ur (a)) = Ur0 (a0 ), a0 = f (a). To
give examples of such maps, we use the standard algebraic structure on Zm .
It is proved (see [78]) that all monomial dynamical systems belong to the
class A(O).
We are interested in attractors of dynamical system (19) (these are ideas-
solutions). To define attractors in the space of ideas XID , we have to define
a convergence in this space. We must introduce a distance on the space
of ideas (sets of associations). Unfortunately there is some mathematical
complication. A metric on the space of points does not induce a metric on
the space of sets which provides an adequate description of the convergence of
ideas. It is more useful to introduce a generalization of the notion of metric,
namely the so called pseudo-metricpseudo-metric In fact, it is even possible
to introduce a metric (Hausdorff’s metric) as people in general topology do,
see [72] and chapter 7. However, it seems that this metric does not give an
adequate description of the dynamics of associations and ideas.. Hence the
dynamics of ideas is a dynamics not in a metric space, but in a more general
space, so called pseudo-metric space.
Let (X, ρ) be a metric space. The distance between a point a ∈ X and a
subset B of X is defined as
ρ(a, B) = inf ρ(a, b)
b∈B

(if B is a finite set then ρ(a, B) = minb∈B ρ(a, b)).

Denote by Sub(X) the system of all subsets of X. Hausdorff distance
between two sets A and B belonging to Sub(X) is defined as
ρ(A, B) = sup ρ(a, B) = sup inf ρ(a, b). (20)
a∈A a∈A b∈B

40
If A and B are finite sets, then

ρ(A, B) = max ρ(a, B) = max min ρ(a, b).

a∈A a∈A b∈B

Hausdorff’s distance ρ is not a metric on the set Y = Sub(X). In particular,

ρ(A, B) = 0 does not imply that A = B. Nevertheless, it can be proved that
the triangle inequality

ρ(A, B) ≤ ρ(A, C) + ρ(C, B), A, B, C ∈ Y,

holds for Hausdorff’s distance.

Let T be a set. A function ρ : T × T → R+ = [0, +∞) for which the
triangle inequality holds true is called a pseudo-metric on T ; (T, ρ) is called
a pseudo-metric space. Hausdorff’s distance is a pseudo-metric on the space
Y of all subsets of the metric space X; (Y, ρ) is a pseudo-metric space. The
strong triangle inequality

ρ(A, B) ≤ max[ρ(A, C), ρ(C, B)] A, B, C ∈ Y,

holds for Hausdorff’s distance corresponding to an ultrametric ρ on X. In

this case Hausdorff’s distance ρ is an ultra-pseudo-metricultra-pseudo-metric
on the set Y = Sub(X).

10 How Can the Brain Play Dice?

Our conscious experience tells us that the deterministic thinking model (17) -
(19), does not provide an adequate description of the real process of thinking.
It seems that a new state of mind depends not only on the previous state
of mind, but also on a (random) choice of a new map f : XI → XI (which
will perform a new iteration), see [68]. What is the basis of such a game
of chance? The contemporary level of neurophysiological research is not
sufficient to obtain a definite answer to this question. We present some
possible mathematical models. Suppose that a cognitive system τ has N
different thinking processors π1 , . . . , πN , with dynamical functions fz , z =
1, 2, . . . , N . The τ uses different blocks for processing of an I-state. At each
instant of time t = 0, 1, . . ., the τ chooses some processor πz and perform a
new iteration:
xn+1 = fz (xn ). (21)
How does τ choose a sequence of processes πz1 , πz2 , . . . , πzn+1 , . . .? The sim-
plest model is a model of the deterministic choice

zn+1 = g(zn ). (22)

41
However, such a system τ will exhibit quite primitive cognitive behaviour.
Example 3.1. Let processors π1 , π2 , π3 , π4 be responsible, respectively,
for visual, articular, sexual and nutritive information. Let g(1) = 2, g(2) =
3, g(3) = 4. Starting with z1 = 1 and x0 = { an image of a pretty women γ},
at the next step τ always uses π2 = πz2 (conversation with this γ), then τ
uses π3 = πz2 (sexual behaviour) and, finally, τ uses π4 = πz4 (interest to
food).
If g(4) = 1 then τ exhibits cyclic behaviour. A τ whose choice mechanism
is ruled by a deterministic law (22) could not change his thinking blocks
depending on the previous I-state xn . For example, if γ is hungry (which
becomes clear after the first two iteration, x2 = fz2 (fz1 (x0 )), it would be
more natural for τ to apply first the block π4 (food) and only after this to
apply the block π3 (sexual behaviour).
The model based on selection rule (22) for dynamical laws describes cog-
nitive behaviour of quite primitive living forms. For example, it could explain
some of the essential features of cognitive behaviour of insects. Higher level
cognitive systems do not just performWe underline that we do not claim that
a choice function g must be represented by a Turing machine. the ‘algorithms’
(22). Their choice essentially depends on the previous I-state xn :

zn+1 = g(zn , xn ). (23)

For example, (23) could explain some essential features of the cognitive be-
haviour of animals. However, (23) could not still explain human cognitive
behaviour. We could not predict with 100% probability that a new state of
a human being τ will be:

xn+1 = fzn+1 (xn ), zn+1 = g(zn , xn )

.
At the next level of complexity τ uses a random selection mechanism:

zn+1 = g(zn , xn , ω), (24)

where ω is a choice parameter. This is a random evolution. Here the implicit

value g(z, x, ω) is not so important. I-dynamics of τ is statistical dynamics:

xn+1 (ω) = fzn+1 (xn (ω)). (25)

Here the value xn+1 (ω) of a new I-state of τ depends on the choice of ω.
The chance parameter ω can also evolve with time: ω = θω, where θ : Ω →
Ω is a law of evolution and Ω is the space of chance parameters. Thus z

42
evolves as z0 , z1 = g(z0 , x0 , ω), z2 = g(z1 , x1 , θω), z3 = g(z2 , x2 , θ2 ω), . . . , see
[15]. Finally we have:
zn+1 (ω) = g(zn (ω), xn (ω), θn−1 ω) , (26)
xn+1 (ω) = fzn+1 (ω) (xn (ω)) . (27)
Roughly speaking τ does not try to find a ‘right decision’ for each triple
(z, x, ω); τ tries only to control its behaviour statistically. Statistical I-
behaviour is determined by probabilities, namely conditional probabilities,
P (xn+1 = y/previous), to obtain at the next step an I-state y on the basis
of information about previous information states, see chapter 4.
One of the distinguishing features of the random dynamics (26), (27) is
that in general such a stochastic process is non-Markovian. We recall that a
stochastic process (chain) {xn (ω)}∞
n=0 has a Markov property if

P (xn+1 = y|xn = k, xn−1 = v, . . . , x0 = λ) = P (xn+1 = y|xn = n). (28)

Here the probability of obtaining a new state xn+1 = y depends only on the
previous state xn = n of the system (and it does not depend on the earlier
evolution x0 = λ, . . . , xn−1 = v). The detailed mathematical investigation
[68], [131], [6] demonstrated that the Markov property of the random evolu-
tion (26), (27) depends strongly on the initial I-state x0 = λ and the structure
of random evolution law θ. For some θ, the I-dynamics is Markovian for any
choice of x0 = λ. Such a cognitive system τ does not use memory on a long
range evolution to create a new I-state xn+1 = y. Here the previous I-state
xn = k determines (but, of course, only statistically) the next state xn+1 = y.
Moreover, some θ (the so called Bernoulli process) induces an I-dynamics
which does not have even one step memory: P (xn+1 = y|xn = n) = P (y),
[68]. Here the randomness of θ is so strong that it destroys even one step
memory. However, the most interesting feature of the dynamics (26), (27) is
that, for a wide class of θ, a τ can demonstrate Markovian as well as non-
Markovian behaviour depending on an initial I-state x0 = λ, [131], [6]. Some
I-states λ are proceeded with one step memory, but others are proceeded
with the long range memory. In the latter case to determine xn+1 = y,
τ uses all the information which was collected in the previous I-evolution,
x0 = λ, x1 = q, . . . , xn−1 = v, xn = w. Another interesting feature of this
model is that the Markovness of the I-evolution depends on the base m of the
coding system: the same flow of chance θ can generate Markovian dynamics
for one tree and non-Markovian dynamics for another.
If, for each z, a map fz belongs to the class A(XI ), then the random
I-dynamics (26), (27) induces the dynamics:
An+1 (ω) = fzn+1 (ω) (An (ω)) (29)

43
of random associations An = An (ω). The dynamics of random associations
(29) automatically induces the dynamics

Jn+1 (ω) = fzn+1 (ω) (Jn (ω)) (30)

of the random ideas, Jn = Jn (ω).

Remark 3.1. (The origin of the game of chance). It is natural to apply
ideas of the standard theory of random dynamical systems [15] and interpret
ω as a noise parameter which disturbs the deterministic I-dynamics of τ .
However, such a viewpoint of a chance parameter ω is too restrictive. In
fact, ω plays some cognitive role. It is not a purely noisy disturbance of
τ . By using ω and its flow θ the τ performs a choice of a dynamical map
fz . It is more natural to assume that ω = (ωnoise , ωcogn ), where ωnoise is a
noisy chance parameter and ωcogn is a random parameter which is specially
produced by τ to perform choice game (26). Such a parameter ωcogn can be
produced by a kind of random generator G. It may be that G is based on
the net of connections between neurons. Such a net has a special statistical
behaviour which is described by a random flow θcogn : Ωcogn → Ωcogn where
Ωcogn is the set of states of this net.

11 Constraints on Information Spaces

It is a general opinion in psychology (see, for example, [158], [96] and see also
[146] for a general discussion) that there are two different types of thinking
processes: (a) associative thinking, (b) goal oriented thinking. According to
[146], p.26, in the first process, one mind (or image), activates another mind
(or image), one image in memory activates another. Such an associative
chain of minds is typical for the situation, in that our mind at some moment
of time has no concrete aim. In the second case we try to solve some problem
or, for instance, to recall the name of some person. Here if the chain of minds
deviates from the main problem (due to some association which appears in
the process of thinking or some external perturbation), then our thinking
system must turn back to the main stream of the aim-oriented thinking. We
shall try to find a counterpart of this cognitive model in our mathemati-
cal model. First, we remark that in our model all thinking processes are
associative processes. Therefore we could not support the viewpoint that
aim-oriented thinking is not based on the flow associations. If a dynamical
thinking processor xn+1 = f (xn ) is described by a continuous function f on
the m-adic tree, then a new I-state is produced on the basis of the nearness
in the sense of a common association to the previous I-state.

44
 However, there is a difference in the functioning of thinking processors
in the regimes (a) and (b). In fact, all our previous considerations were
related to the regime (a). To describe (b), we need to modify the scheme
which was used in previous sections. We shall use the framework of random
dynamical systems. We restrict ourselves to the dynamics of I-states. A
new I-state is produced via the random dynamics (26), (27). Let the aim
of the process of thinking be represented by an I-state aaim . In the process
(b) of aim-oriented thinking each new I-state xn+1 (ω) must be compared
with the aim aaim : ρm (xn+1 , aaim ) ≤ , where  > 0 gives the sharpness
of aim-oriented thinking. The thinking trajectory x0 , x1 , . . . , xn , xn+1 , . . .
must belong to the neighbourhood U (aaim ) = {x : ρm (x, aaim ) ≤ } of the
aim. Suppose now that at some step this trajectory leaves the neighbour-
hood: ρm (xn+1 , aaim ) > . At this moment the trajectory must be corrected:
x0n+1 (ω) = fzn+1 (ω) (xn (ω)), must be deleted and iterations start again with
the previous I-state xn (which still belongs U (aaim )). However, we do not
obtain xn+1 , again because the chance flow θ continues its stream. A new
I-state x0n+1 is obtained as

x0n+1 (ω) = fzn+2 (ω) (xn (ω)),

where
zn+2 (ω) = g(zn+1 (ω), xn+1 (ω), θn ω) .
This new I-state may belong to U (aaim ). In such a case it is accepted as
the right iteration in aim-oriented thinking. If x0n+1 (ω) 6∈ U (aaim ), then τ
computes
x00n+1 (ω) = fzn+3 (ω) (xn (ω))
and so on.
[p-adic Dynamical Systems] p-adic Dynamical Systems
The fields Qp of p-adic numbers (where p = 2, 3, . . . , 1999, . . . are prime
numbers) were introduced by the German mathematician K. Hensel at the
end of 19th century [91]. Hensel started with the following question:
Is it possible to expand a rational number x ∈ Q in a power series of the
form ∞
X
x= αn pn , αn = 0, . . . , p − 1, (31)
n=k

where k = 0, ±1, ±2, . . . depends on x. Of course, this question was moti-

vated by the existence of real expansions of rational numbers with respect to
a p-scale:
X k
x= αn pn , αn = 0, . . . , p − 1 . (32)
n=−∞

45
Hensel knew, for example, that
0
4 X
= 22n .
3 n=−∞

4
He studied the possibility of expanding x = 3
in a series with respect to
positive powers of p = 2 :
∞
4 X
= αn 2n , αn = 0, 1.
3 n=0

Such rather innocent manipulations with rational numbers and series gener-
ated the idea that there exists some algebraic structure similar to the system
of real numbers R. Hensel observed that it is possible to introduce algebraic
operations (addition, subtraction, multiplication, division) on the set Qp of
all formal series (31). Thus each Qp has the structure of a number field. The
field of rational numbers Q is a subfield of Qp . In fact, the fields of p-adic
numbers Qp were first examples of infinite fields that differs from Q, R, C
and fields of rational functions.
As we have already mentioned, it is possible to work in a more general
framework, namely, to consider not only prime numbers p, but all natural
numbers m as bases of expansions. In principle we can do this. However,
the corresponding number system is not in general a field. If m = p1 p2 ,
where pj are distinct prime numbers, then Qm is not a field (there exist
divisors of zero) but only a ring, i.e., division is not well defined. The field
structure is very important for developing analysis. Therefore the main part
of investigations have been performed for prime p.
The construction of new fields Qp induced strong interest in number the-
ory and algebra. For practically one hundred years p-adic numbers were
intensively used only in pure mathematics, mainly in number theory; see,
for example, the classical book of Borevich and Schafarevich [40]. In partic-
ular, p-adic numbers are useful in investigations of some number-theoretic
problems in the field Q of rational numbers. Typically, if we can prove some
result for the field of real numbers R as well as for all fields of p-adic num-
bers Qp , p = 2, 3, . . . , then we obtain the corresponding result for the field of
rational numbers Q; see [40] for the details.
The presence of the field structure on Qp gives the possibility of develop-
ing analysis for functions f : Qp → Qp . In particular, the derivativederivative
of such a function is defined in the usual way:
f (x + h) − f (x)
f 0 (x) = lim , x, h ∈ Qp .
h→0 h

46
 Of course, to perform a limit procedure we need a topology on Qp . It is
possible to define on Qp a p-adic absolute value, valuation, x → |x|p that has
properties similar to the properties of the ordinary absolute value on R. The
topology on the field Qp is defined by the metric ρp (x, y) = |x − y|p . The Qp
is a locally compact topological field (i.e., the unit ball of Qp is a compact set
and all algebraic operations are continuous).
The important event in the p-adic world took place in 1987 when I.
Volovich published the first paper [190] on applications of p-adic numbers
in theoretical physics. This paper was devoted to the p-adic string the-
ory.string theory The string theory was a new and rather intriguing attempt
to reconsider the foundations of physics by using spatially extended objects,
strings, instead of the point objects, elementary particles. The scenarios of
the string spectacle is performed at fantastically small distances, so called
Planck distances, lP ≈ 10−34 cm. Physicists have (at least) the feeling that
space–time at Planck distances has some distinguishing features which could
not be described by the standard mathematical model based on the field of
real numbers R. In particular, there are ideas (which also are strongly moti-
vated by cosmology We remark that one of the aims of string theory was to
provide a new approach to general relativity. Therefore string investigations
are closely connected to investigations in the fields of gravity and cosmology.)
that at Planck distances we can no longer assume that there is a kind of an
order structure on the real line R. We remark that there is no order structure
on Qp (this is a disordered field).
Another argument for considering a p-adic model of space–time at Planck
distancesPlanck distances is that Qp is a non-Archimedeannon-Archimedean
field. We do not plan to discuss here the Archimedean axiom at the mathe-
matical level of rigorousness.
From the physical point of view this axiom can be interpreted in the
following way. If we have some unit of measurement l then we can measure
each interval L by using l. By addition of this unit, l, l+l, l+l+l, . . . , l+. . .+l,
we obtain larger and larger intervals which, eventually, will cover L. The
precision of such a measurement is equal to l. The process of such a type
can be realized in the field of real numbers R. Therefore all physical models
based on real numbers are Archimedean models. However, the Archimedean
axiom does not hold true in Qp . Here successive addition does not increase the
quantity. And there were (long before p-adic physics) intuitive cosmological
ideas that space–time at Planck distances has a non-Archimedean structure.
In any case the paper of Volovich [190] induced a large interest in p-
adic physical models; see, for example, papers on p-adic string theory by
Aref’eva, Brekke, Dragovich, Framton, Freud, Parisi, Vladimirov, Volovich,
Witten, and many others referred to in chapter 1. These p-adic physical

47
investigations stimulated the large interest in dynamical systems in fields of
p-adic numbers Qp and their finite and infinite extensions (and, in particular,
in the field of complex p-adic numbers Cp ).
Investigations into p-adic (discrete) dynamical systems is the best illustra-
tion of how physical models can stimulate new mathematical investigations.
Starting with the papers on p-adic quantum mechanics and string theory,
investigations into discrete p-adic dynamical systems were carried out in var-
ious directions, see references in chapter 1.

12 p-Adic Numbers
The field of real numbers R is constructed as the completion of the field of
rational numbers Q with respect to the metric ρ(x, y) = |x − y|, where | · | is
the usual valuation given by the absolute value. The fields of p-adic numbers
Qp are constructed in a corresponding way, but using other valuations. For a
prime number p the p-adic valuation |·|p is defined in the following way. First
we define it for natural numbers. Every natural number n can be represented
as a product of prime numbers, n = 2r2 3r3 · · · prp · · · , and we define |n|p =
p−rp , writing |0|p = 0 and | − n|p = |n|p . We then extend the definition of the
p-adic valuation | · |p to all rational numbers by setting |n/m|p = |n|p /|m|p
for m 6= 0. The completion of Q with respect to the metric ρp (x, y) = |x−y|p
is the locally compact field of p-adic numbers Qp .
The number fields R and Qp are unique in a sense, since by Ostrovsky’s
theoremOstrovsky’s theorem [40] | · | and | · |p are the only possible valuations
on Q, but have quite distinctive properties. The field of real numbers R with
its usual valuation satisfies |n| = n → ∞ for valuations of natural numbers n
and is said to be Archimedean. By a well known theorem of number theory
[163] the only complete Archimedean fields are those of the real and the
complex numbers. In contrast, the fields of p-adic numbers which satisfy
|n|p ≤ 1 for all n ∈ N are examples of non-Archimedean fields. Here the
Archimedean axiom is violated. We could not obtain a larger quantity by
successive addition. Let l be any element of Qp such that |l|p ≤ 1. There
does not exist such a natural number n that |nl|p ≤ 1.
The field of real numbers R is not isomorphic to any Qp . Fields Qs and Qt
are not isomorphic for s 6= t. Thus starting with the field of rational numbers
Q we obtain an infinite series of locally compact non-isomorphic fields:

Q2 , Q3 , ..., Q1997 , Q1999 , ..., Q213466917 −1 , ... .

Any x ∈ Qp has a unique canonical expansion (which converges in the | · |p –

48
norm) of the form

x = a−n /pn + · · · a0 + · · · + ak pk + · · · , (33)

where the aj ∈ {0, 1, . . . , p − 1} are the ‘digits’ of the p-adic expansion and
n depend on x (cf. (1.7) in chapter 1, where we introduced m-adic numbers
as a set of such power series).
We remark that because Qp is a locally compact additive group there
exists the Haar measureHaar measure dx on the σ-algebra of Borel subsets
of Qp .
As we have already mentioned, the derivative of a function f : Qp → Qp
is defined (as usual) as
f (x + ∆x) − f (x)
f 0 (x) = lim .
|∆x|p →0 ∆x
The main distinguishing feature of p-adic analysis is the existence of non-
locally constant functions with zero derivative. We present the following well
known example [163], p.74. The function f : Zp → Zp is defined as
∞
X ∞
X
f (x) = an p2n for x = an p n .
n=0 n=0

This function is injective (f (x1 ) 6= f (x2 ) for x1 6= x2 ) and f 0 ≡ 0.

The p-adic exponential functionexponential function is defined in the
usual way by using power series expansion:
∞
x
X xn
e = .
n=0
n!

In the p-adic case the exponential function is not entire analytic. The series
converges in Qp if

|x|p ≤ rp , where rp = 1/p, p 6= 2, and r2 = 1/4. (34)

p-adic trigonometric functions sin x and cos x are defined by the standard
power series. These series have the same radius of convergence rp as the
exponential series.
We can consider a p-adic analogue of complex numbers. As we know, the
field of complex numbers C is the quadratic extension
√ of R with respect to the
2
root of the equation x + 1 = 0 : C = R(i), i = −1, z = x + iy, x, y ∈ R.
In this case we have a very simple algebraic structure, because this quadratic
extension is at the same time the algebraic closure of the field of real numbers

49
(every polynomial equation has a solution in C). In the p-adic case the
structure of algebraic extensions is more complicated. A quadratic extension
is not unique. If p = 2 then there are seven quadratic extensions, and if
p 6= 2 then there are three quadratic
√ extensions. Thus if we consider the
fixed quadratic extension Qp ( τ ) of Qp then there exist √ p-adic numbers
for which it is impossible to find a square root in Qp ( τ ). All quadratic
extensions are not algebraically closed. Extensions of any finite order (i.e,
corresponding to roots of polynomials of any order) are not algebraically
closed. The algebraic closure Qap of Qp is constructed as an infinite chain of
extensions of finite orders. Therefore this is an infinite-dimensional vector
space over Qp . This algebraic closure is not a complete field. Thus we must
consider the completion of this field. It is the final step of this long procedure,
because this completion is an algebraically closed field (so we are lucky!),
Krasner’s theorem, see, e.g., [163]. Let us denote this field by Cp . This field
is called the field of complex p-adic numbers.

13 Roots of Unity
The roots of unity in Cp will play the important role in our considerations.
To find fixed points and cycles of monomial functions ψn (x) = xn , we have
to find the roots of unity.
As usual in arithmetics, (n, k) denotes the greatest common divisor of
two natural numbers. Denote the group of mth roots of unity, m = 1, 2, ...,
by Γ(m) . Set
j
Γ = ∪∞
m=1 Γ
(m)
, Γm = ∪∞
j=1 Γ
(m )
, Γu = ∪(m,p)=1 Γm ,

By elementary group theory we have Γ = Γu · Γp , Γu ∩ Γp = {1}. In chapter

1 we denoted the kth roots of unity by θj,k , j = 1, ..., k, and chose θ1,k = 1.
We remark that Γu ⊂ S1 (1) and Γp ⊂ U1− (1), [9].
The following estimate plays the important role in p-adic analysis and
theory of monomial dynamical systems. We also present the proof to demon-
strate the rules of working in the framework of p-adic analysis.
Lemma 2. 1. | pjk |p ≤ 1/p for all j = 1, ..., pk − 1.

Proof. Let j = ip + q, q = 0, 1, ..., p − 1. First consider the case q = 0 :

k k k −ip+p)
j
|p = | p (p −p)···(p


| pk p···ip
|p
 k k   k−1 
 p (p −p) (pk −ip+p) 
=  ip p · · · ip−p  =  p i  ≤ p1 ,
 
p p

50
as i < pk−1 . Now let q 6= 0 :
 k (pk − p) (pk − ip) 
   
j 1
| k |p = p  ··· = |pk |p ≤ .
p p ip 
p p

To find fixed points and cycles of functions ψn (x) = xn in Qp we have to

know whether the roots of unity belong to Qp . We present the corresponding
result. Denote by ξl , l = 1, 2, ..., a primitive lth root of 1 in Cp . We are
interested in whether ξl ∈ Qp .
Proposition 2. 1. (Primitive roots) If p 6= 2 then ξl ∈ Qp if and only if
l | (p − 1). The field Q2 contains only ξ1 = 1 and ξ2 = −1.
To prove this proposition we have to prove the same result for the field
Fp = {0, 1, ..., p − 1} of mod p residue classes and apply Hensel’s lemma [40],
[163], [107] (p-adic variant of Newton method, see section 7). This is one of
the most powerful methods for obtaining results for Qp : first obtain such
a result for Fp and try to find conditions for applying Hensel’s lemma. In
particular, we obtain the following result, see chapter 1, Proposition 5.1.
Corollary 2.1. The equation xk = 1 has g = (k, p − 1) different roots
in Qp .

14 Dynamical Systems in Non-Archimedean Fields

To study dynamical systems in fields of p-adic numbers Qp and complex p-
adic numbers Cp as well as finite extensions of Qp , it is convenient to consider
the general case of an arbitrary non-Archimedean field K.
Let K be a field (so all algebraic operations are well defined). Recall
that a non-Archimedean valuation is a mapping | · |K : K → R+ satisfying
the following conditions: |x|K = 0 ⇐⇒ x = 0 and |1|K = 1; |xy|K =
|x|K |y|K ; |x + y|K ≤ max(|x|K , |y|K ). The latter inequality is the well known
strong triangle axiom.strong triangle axiom
The field K with the valuation | · |K is called a non-Archimedean field.
The fields of p-adic numbers Qp and complex p-adic numbers Cp , as well as
finite extensions of Qp are non-Archimedean fields.
We often use in non-Archimedean investigations the following property
of a non-Archimedean valuation:

|x + y|F = max(|x|F , |y|F ) if |x|F 6= |y|F .

Thus all triangles in a non-Archimedean fields (in particular, in fields of

p-adic numbers) are isosceles.

51
 We also remark that every non-Archimedean field K is an ultrametric
space with ultrametric ρK (x, y) = |x − y|K .
Everywhere below K denotes a complete non-Archimedean field with a
nontrivial valuation | · |K ; Ur (a), Ur− (a) and Sr (a) are respectively balls and
spheres in K. We always consider

r ∈ |K| = {s = |x|K : x ∈ K}

for radii of balls Ur (a) and spheres Sr (a). In particular, in the p-adic case
r = pl , l = 0, ±1, ±2, ... and in the case of Cp – r = pq , q ∈ Q.
A function f : Ur (a) → K is said to be analyticanalytic if it can be
expanded into a power series
∞
X
f (x) = fn (x − a)n
n=0

with fn ∈ K which converges uniformly on the ball Ur (a). This is equivalent

to the condition
lim |fn |K rn = 0
n→∞
P
(because, as a consequence of the strong triangle inequality, any series n an , an ∈
K, converges iff an → 0, n → ∞).
Let us study the dynamical system:

U → U, x → f (x), (35)

where U = UR (a) or K and f : U → U is an analytic function. First we shall

prove a general theorem about behaviour of iterations xn = f n (x0 ), x0 ∈ U.
As usual, f n (x) = f ◦ ... ◦ f (x). Then we shall use this result to study a
behaviour of the concrete dynamical systems ψn (x) = xn , n = 2, 3, ..., in the
fields of complex p-adic numbers Cp .
We shall use the standard terminology of the theory of dynamical systems,
see section 4, chapter 1.
Remark 3.1. In complex geometry the center of a disk is uniquely determined
by the disk. Hence it does not happen that different fixed points have the same
Siegel disk. But in non-Archimedean geometry the centers of a disk are nothing
but the points which belong to the disk. And in principle different fixed points
may have the same Siegel disk (see the next section).
As usual in the theory of dynamical systems, we can find attractors,
repellers, and Siegel disks by using properties of the derivative of f (x). Let
a be a periodic point with period n of C 1 -function g : U → U. Set λ =
dg n (a)/dx. This point is called:

52
 1) attractive if 0 ≤ |λ|K < 1;
2) indifferent if |λ|K = 1;
3) repelling if |λ|K > 1.
Lemma 3.1. [111] Let f : U → U be an analytic function and let a ∈ U
and f 0 (a) 6= 0. Then there exist r > 0 such that
 1 dn f  n−1
 
s = max   (a) r < |f 0 (a)|K . (36)
2≤n<∞ n! dxn
K

If r > 0 satisfies this inequality and Ur (a) ⊂ U then

|f (x) − f (y)|K = |f 0 (a)|K |x − y|K (37)
for all x, y ∈ Ur (a).
By using the previous Lemma we prove:
Theorem 3.1. Let a be a fixed point of the analytic function f : U → U.
Then:
1. If a is an attracting point of f then it is an attractor of the dynamical
system (35). If r > 0 satisfies the inequality:
 1 dn f  n−1
 
q = max   (a) r < 1, (38)
1≤n<∞ n! dxn
K

and Ur (a) ⊂ U then Ur (a) ⊂ A(a).

2. If a is an indifferent point of f then it is the center of a Siegel disk. If
r > 0 satisfies the inequality (36) and Ur (a) ⊂ U then Ur (a) ⊂ SI(a).
3. If a is a repelling point of f then a is a repeller of the dynamical system
(35).
We note that (in the case of an attracting point) the condition (38) is less
restrictive than the condition (36).
To study dynamical systems for nonanalytic functions we can use the
following theorem of non-Archimedean analysis [163]:
Theorem 3.2. (Local injectivity of C 1 -functions) Let f : Ur (a) → K be
C 1 at the point a. If f 0 (a) 6= 0 there is a ball Us (a), s ≤ r, such that (37)
holds for all x, y ∈ Us (a).
However, Theorem 3.1 is more useful for our considerations because Theo-
rem 3.2 is a so called ‘existence theorem’. This theorem does not say anything
about the value of s. Thus we cannot estimate a volume of A(a) or SI(a).
Theorem 3.1 gives us such a possibility. We need only to test one of the
conditions (38) or (36).
Moreover, the case f 0 (a) = 0 is ‘a pathological case’ for nonanalytic func-
tions of a non-Archimedean argument. For example, there exist functions g
which are not locally constant but g 0 = 0 in every point.

53
 In our analytic framework we have no such problems.
A Julia setJulia set Jf for the dynamical system (35) is defined as the
closure of the set of all repelling periodic points of f. The set Ff = U \ Jf is
called a Fatou set.Fatou set These sets play an important role in the theory
of real dynamical systems.
We shall also use an analogue of Theorem 3.1 for periodic points. There
we must apply our theorem to the iterated function f n (x).

15 Dynamical Systems in the Field of Complex

p-adic Numbers
As an application of Theorem 3.1 we study the simplest discrete dynamical
systems, namely, monomial systems:
f (x) = ψn (x) = xn , n = 2, 3, ...,
in fields of complex p-adic numbers Cp . We shall see that the behaviour of
ψn crucially depends on the prime number p, the base of the corresponding
field. Depending on what the value p is, attractors and Siegel disks ap-
pear and disappear, transforming one into another. Especially will complex
dependence on p be studied in section 5 devoted to dynamical systems in
Qp , p = 2, 3, ..., 1997, 1999, ... .
It is evident that the points a0 and a∞ are attractors with basins of
attraction A(0) = U1− (0) and A(∞) = Cp \U1 (0), respectively. Thus the main
scenario is developed on the sphere S1 (0). Fixed points of ψn (x) belonging
to this sphere are the roots θj,n−1 , j = 1, ..., n − 1, of unity of degree (n − 1).
There are two essentially different cases: 1) n is not divisible by p; 2) n is
divisible by p. The proof of the following theorem is based on the results of
Theorem 3.1.
Theorem 4.1. The dynamical system ψn (x) has (n − 1) fixed points
aj = θj,n−1 , j = 1, ..., n − 1, on the sphere S1 (0).
1. Let (n, p) = 1. There all these points are centers of Siegel disks and
SI(aj ) = U1− (aj ). If n − 1 = pl , l = 1, 2, ..., then SI(aj ) = SI(1) = U1− (1)
for all j = 1, ..., n − 1. If (n − 1, p) = 1 then aj ∈ S1 (1), j = 2, ..., n − 1, and
SI(aj ) ∩ SI(ai ) = ∅, i 6= j. For any k = 2, 3, ... all k-cycles are also centers
of Siegel disks of unit radius.
2. If (n, p) 6= 1 then these points are attractors and U1− (aj ) ⊂ A(aj ).
For any k = 2, 3, ... all k-cycles are also attractors and open unit balls are
contained in basins of attraction.
Corollary 4.1. Let (n, p) = 1. Let nk − 1 = pl , l = 1, 2, ... Any k-
cycle γ = (a1 , ..., ak ) for such a k is located in the ball U1− (1); it has the

54
behaviour of a Siegel disk with SI(γ) = ∪kj=1 U1− (aj ) = U1− (1). During the
process of the motion the distances cj = ρp (x0 , aj ), j = 1, ..., k, where x0 ∈
U1− (1) is an arbitrary initial point, are changed according to the cyclic law:
(c1 , c2 , ..., cn−1 , cn ) → (cn , c1 , ..., cn−2 , cn−1 ) → ... .
Thus in the case (n, p) = 1 the motion of a point in the ball U1− (1) is very
complicated. It moves cyclically (with different periods) around an infinite
number of centers. Unfortunately, we could not paint a picture of such a
motion in our Euclidean space — it is too restrictive for such images.
Theorem 4.1 does not completely describe the case (n, p) = 1, (n − 1, p) 6=
1. Let us consider the general case: (n, p) = 1, n − 1 = mpl with (m, p) = 1
and l ≥ 0. Set ai = ξm i
, i = 0, 1, ..., m − 1, bj = ξpjl , j = 0, 1, ..., pl − 1, and
cij = ai bj . Then all these points cij , i = 0, 1, ..., m − 1, j = 0, 1, ..., pl − 1, are
centers of Siegel disks and SI(cij ) = U1− (cij ). For each i we have SI(ci0 ) =
SI(ci1 ) = ... = SI(ci(pl −1) ). If i 6= 0 then all these disks SI(cij ) are in
S1 (0) ∩ S1 (1). Further more, SI(cij ) ∩ SI(ckl ) = ∅ if i 6= k. We can formulate
the same result for k-cycles.
Now we find the basins of attraction A(aj ), j = 1, ..., n − 1, (n, p) 6= 1,
exactly. We begin from the attractor a1 = 1.
Lemma 4.1. Let n = mpk , (m, p) = 1, and k ≥ 1. The basin of attrac-
tion A(1) = ∪ξ U1− (ξ) where ξ ∈ Γm ; these balls have empty intersections for
different points ξ.
Theorem 4.2. Let n = mpk , (m, p) = 1, and k ≥ 1. The basin of attrac-
tion A(ak ) = ∪ξ U1− (ξak ) where ξ ∈ Γm . These balls have empty intersections
for different points ξ.

16 Dynamical Systems in the Fields of p-adic

Numbers
Here we study the behaviour of the dynamical system

ψn (x) = xn , n = 2, 3, ...,

in Qp . We start with the remark that the basic Theorem 5.1, chapter 1, can
be obtained on the basis of results obtained in previous section for Cp , see
[111] for the details.
We remark that the number of attractors or Siegel disks for the dynamical
system ψn (x) on the sphere S1 (0) is ≤ (p − 1). To study k-cycles in Qp we
use the following numbers:
mk = (lk , p − 1), k = 1, 2, ..., with lk = nk − 1.

55
 Proposition 5.1. The dynamical system ψn (x) has k-cycles (k ≥ 2) in
Qp if and only if mk does not divide any mj , j = 1, ..., k − 1. All these cycles
are located on S1 (1).
In particular, if (n, p) = 1 (i.e., all fixed points and k-cycles are centers
of Siegel disks) there are no such complicated motions around a group of
centers as in Cp .
Corollary 5.1. The dynamical system ψn (x) has only a finite number of
cycles in Qp for any prime number p.
We now consider prime number dependence of cyclic behaviour for the dy-
namical system: ψ2 (x) = x2 . This is the simplest amongst monomial dynam-
ical systems. However, even here we observe very complicated dependence
on p.
Examples. Let n = 2 in all the following examples.
1). Let p = 2. There is only one fixed point a1 = 1 on S1 (0). It is an
attractor and A(1) = U1/2 (1) = S1 (0).
2). As lk are odd numbers, mk must also be an odd number. Therefore
there are no k-cycles (k > 1) for p = 3, 5, 17 nor for any prime number which
has the form p = 2k + 1.
3). Let p = 7. Here mk can be equal to 1 or 3. As m2 = 3 there are only
2-cycles. It is easy to show that the 2-cycle is unique.
4). Let p = 11. Here mk = 1 or 5. As m2 = m3 = 1 and m4 = 5 there
exist only 4-cycles. There is only one 4-cycle: γ(ξ5 ).
5). Let p = 13. Here mk = 1 or 3. As m2 = 3 there exists only the
(unique) 2-cycle.
6). Let p = 19. Here mk = 1 or 3, or 9. As m2 = 3 there is the (unique)
2-cycle. However, although m4 = 3 there are no 4-cycles because m4 divides
m3 . As m6 = 9 does not divide m2 , ..., m5 there exist 6-cycles and there are
no k-cycles with k > 6. There is only one 6-cycle: γ(ξ9 ).
7). Let p = 23. Here mk = 1 or 11. The direct computations show
that there are no k-cycles for the first k = 2, ..., 8. Further computations are
complicated.
8). Let p = 29. Here mk = 1 or 7. As m3 = 7 and m2 = 1 there exist
only 3-cycles. It is easy to show that there are two 3-cycles: γ(ξ7 ) and γ(ξ73 ).
9). Let p = 31. Here mk = 1, 3, 5, 15. As m2 = 3 there exists an (unique)
3 7
2-cycle. As m4 = 15 and m3 = 1 there exist 4-cycles: γ(ξ15 ), γ(ξ15 ), γ(ξ15 ).
There are no k-cycles with k 6= 2, 4.
10). Let p = 37. Here mk = 1, 3, 9. As m2 = 3 there exists an (unique)
2-cycle. As m6 = 9 and m2 = m4 = 3, m3 = m5 = 1 there exist 6-cycles. It
is easy to show that there is an unique 6-cycle: γ(ξ9 ). There are no k-cycles
for k 6= 2, 6.

56
 11). Let p = 41. Here mk = 1, 5. As m4 = 5 and all previous mj = 1
there exist 4-cycles. It is easy to show that this cycle is unique: γ(ξ5 ). There
are no k-cycles with k 6= 4.
Thus, even for the simplest p-adic dynamical system, x → x2 , the struc-
ture of cycles depends in a very complex way on the parameter p. In the
general case this dependence was studied in [122], [145]. It was demon-
strated that a number Nm (p) of cycles of the fixed length m in Qp depends
randomly on p. So Nm (p) is a random variable defined on the set of all prime
numbers. We found the mean value and covariance of this random variable
[122], [145]. There we used essentially classical results on the distribution
of prime numbers, see, e.g., [40]. We hope that the connection between the
theory of p-adic dynamical systems and the classical theory of distributions
of prime numbers established in [122] will have further applications.

17 p-adic Ergodicity
In this section we study in detail the ergodic behavior of p-adic monomial
dynamical systems. As we have already seen in the previous section, the
behaviour of p-adic dynamical systems depends crucially on the prime pa-
rameter p. The main aim of investigations performed in the paper [86] was
to find such a p-dependence for ergodicity.
Let ψn be a (monomial) mapping on Zp taking x to xn . Then all spheres
Sp−l (1) are ψn -invariant iff n is a multiplicative unit, i.e., (n, p) = 1.
In particular, ψn is an isometry on Sp−l (1) if and only if (n, p) = 1.
Therefore we will henceforth assume that n is a unit. Also note that as a
consequence Sp−l (1) is not a group under multiplication. Thus our investi-
gations are not about the dynamics on a compact (abelian) group.
We remark that monomial mappings, x 7→ xn , are topologically transitive
and ergodic with respect to Haar measure on the unit circle in the complex
plane. We obtained [86] an analogous result for monomial dynamical systems
over p-adic numbers. The process is, however, not straightforward. The
result will depend on the natural number n. Moreover, in the p-adic case we
never have ergodicity on the unit circle, but on the circles around the point
1.
6.1. Minimality.
Let us consider the dynamical system x 7→ xn on spheres Sp−l (1). The
result depends crucially on the following well known result from group theory.
We denote the multiplicative group of the ring Fk of mod k residue classes
by the symbol Fk? ; we also set < n >= {nN : N = 0, 1, 2, ...} for a natural
number n.

57
 Lemma 6.1. Let p > 2 and l be any natural number, then the natural
number n is a generator of Fp?l if and only if n is a generator of Fp?2 . F2?l is
noncyclic for l ≥ 3.
Recall that a dynamical system given by a continuous transformation ψ on
a compact metric space X is called topologically transitive if there exists a
dense orbit {ψ n (x) : n ∈ N} in X, and (one-sided) minimal if all orbits for
ψ in X are dense. For the case of monomial systems x 7→ xn on spheres
Sp−l (1) topological transitivity means the existence of an x ∈ Sp−l (1) s.t.
each y ∈ Sp−l (1) is a limit point in the orbit of x, i.e., can be represented as
Nk
y = lim xn (39)
k→∞

for some sequence {Nk }, whilst minimality means that such a property holds
for any x ∈ Sp−l (1). Our investigations are based on the following theorem.
Theorem 6.1. For p 6= 2 the set hni is dense in S1 (0) if and only if n is
a generator of Fp?2 .
Proof. We have to show that for every  > 0 and every x ∈ S1 (0)
there is a y ∈ hni such that |x − y|p < . Let  > 0 and x ∈ S1 (0) be
arbitrary. Because of the discreteness of the p-adic metric we can assume
that  = p−k for some natural number k. But (according to Lemma 6.1) if
n is a generator of Fp?2 then n is also a generator of Fp?l for every natural
number l (and p 6= 2) and especially for l = k. Consequently there is an N
such that nN = x mod pk . From the definition of the p-adic metric we see
that |x − y|p < p−k if and only if x equals to y mod pk . Hence we have that
 
x − nN  < p−k .
p
Let us consider p 6= 2 and for x ∈ Up−1 (1) the p-adic exponential function
t 7→ xt , see, for example [163]. This function is well defined and continuous
as a map from Zp to Zp . In particular, for each a ∈ Zp we have
xa = lim xk , k ∈ N. (40)
k→a

We shall also use properties of the p-adic logarithmic function; see, for ex-
ample [163]. Let z ∈ Up−1 (1). Then log z is well defined. For z = 1 + λ with
|λ|p ≤ 1/p we have:
∞
X (−1)k+1 λk
log z = = λ(1 + λ∆λ ), |∆λ |p ≤ 1. (41)
k=1
k

By using (41) we obtain that log : Up−1 (1) → Up−1 (0) is an isometry:
| log x1 − log x2 |p = |x1 − x2 |p , x1 , x2 ∈ U1/p (1) . (42)

58
 Lemma 6.2. Let x ∈ Up−1 (1), x 6= 1, a ∈ Zp and let {mk } be a sequence
of natural numbers. If xmk → xa , k → ∞, then mk → a as k → ∞, in Zp .
This is a consequence of the isometric property of log.
Theorem 6.2. Let p 6= 2 and l ≥ 1. Then the monomial dynamical
system x 7→ xn is minimal on the circle Sp−l (1) if and only if n is a generator
of Fp?2 .
Proof. Let x ∈ Sp−l (1). Consider the equation xa = y. What are
the possible values of a for y ∈ Sp−l (1)? We prove that a can take an
arbitrary value from the sphere S1 (0). We have that a = log x/ log y. As
log : Up−1 (1) → Up−1 (0) is an isometry, we have log(Sp−l (1)) = Sp−l (1). Thus
a = log x/ log y ∈ S1 (0) and, moreover, each a ∈ S1 (0) can be represented as
log x/ log y for some y ∈ Sp−l (1).
Let y be an arbitrary element of Sp−l (1) and let xa = y for some a ∈ S1 (0).
By Theorem 6.1, if n is a generator of Fp?2 then each a ∈ S1 (0) is a limit point
of the sequence {nN }∞ N =1 . Thus a = limk→∞ n
Nk
for some subsequence {Nk }.
By using the continuity of the exponential function we obtain (39).
N
Suppose now that xn k → xa for some n. By Lemma 6.2 we obtain
that nNk → a as k → ∞. If we have (39) for all y ∈ Sp−l (1) then each
a ∈ S1 (0) can be approximated by elements nN . In particular, all elements
{1, 2, ..., p − 1, p + 1, .., p2 − 1} can be approximated with respect to mod p2 .
Thus n is a is a generator of Fp?2 .
Example 6.1. In the case that p = 3 we have that ψn is minimal if n = 2,
2 is a generator of F9? = {1, 2, 4, 5, 7, 8}. But for n = 4 it is not; h4imod 32 =
{1, 4, 7}. We can also see this by noting that S1/3 (1) = U1/3 (4) ∪ U1/3 (7) and
that U1/3 (4) is invariant under ψ4 .
Corollary 6.1. If a is a fixed point of the monomial dynamical system
x 7→ xn , then this is minimal on Sp−l (a) if and only if n is a generator of
Fp?2 .
Example 6.2. Let p = 17 and n = 3. In Q17 there is a primitive 3rd
?
root of unity, see for example [163]. Moreover, 3 is also a generator of F17 2.

Therefore there exist nth roots of unity different from 1 around which the
dynamics is minimal.
6.2. Unique ergodicity.
In the following we will show that the minimality of the monomial dy-
namical system ψn : x 7→ xn on the sphere Sp−l (1) is equivalent to its unique
ergodicity. The latter property means that there exists a unique probability
measure on Sp−l (1) and its Borel σ−algebra which is invariant under ψn .
We will see that this measure is, in fact, the normalized restriction of the
Haar measure on Zp . Moreover, we will also see that the ergodicity of ψn

59
with respect to the Haar measure is also equivalent to its unique ergodicity.
We should point out that although many results are analogous to the case
of the (irrational) rotation on the circle, our situation is quite different, in
particular as we do not deal with dynamics on topological subgroups.
Lemma 6.3. Assume that ψn is minimal. Then the Haar measure m is
the unique ψn -invariant measure on Sp−l (1).
Proof. First note that minimality of ψn implies that (n, p) = 1
and hence that ψn is an isometry on Sp−l (1). Then, as a consequence of
Theorem 27.5 in [163] it follows that ψn (Ur (a)) = Ur (ψn (a)) for each ball
Ur (a) ⊂ Sp−l (1). Consequently for every open set U 6= ∅ we have Sp−l (1) =
∪∞ N
N =0 ψn (U ). It follows for a ψn −invariant measure µ that µ(U ) > 0.
Moreover, we can split Sp−l (1) into disjoint balls of radii p−(l+k) , k ≥ 1,
on which ψn acts as a permutation. In fact, for each k ≥ 1, Sp−l (1) is the
union,
Sp−l (1) = ∪Up−(l+k) (1 + bl pl + ... + bl+k−1 pl+k−1 ), (43)
where bi ∈ {0, 1, ..., p − 1} and bl 6= 0.
We now show that ψn is a permutation on the partition (43). Recall that
every element of a p-adic ball is the center of that ball, and as pointed out
above ψn (Ur (a)) = Ur (ψn (a)). Consequently for all positive integers k we
have ψnk (a) ∈ Ur (a) ⇒ ψnk (Ur (a)) = Ur (ψnk (a)) = Ur (a), so that ψnN k (a) ∈
Ur (a) for every natural number N . Hence for a minimal ψn a point of a ball
B of the partition (43) must move to another ball in the partition.
Furthermore, the minimality of ψn shows indeed that ψn acts as a permu-
tation on balls. By the invariance of µ all balls must have the same positive
measure. As this holds for any k, µ must be the restriction of the Haar
measure m.
The arguments of the proof of Lemma 6.3 also show that the Haar measure
is always ψn -invariant. Thus if ψn is uniquely ergodic the unique invariant
measure must be the Haar measure m. Under these circumstances it is known
that ψn must be minimal.
Theorem 6.3. The monomial dynamical system ψn : x 7→ xn on Sp−l (1)
is minimal if and only if it is uniquely ergodic in which case the unique
invariant measure is the Haar measure.
Let us mention that unique ergodicity yields in particular the ergodicity of
the unique invariant measure, i.e., the Haar measure dx, which means that
N −1 Z
1 X ni
f (x ) → f dx for all x ∈ Sp−l (1), (44)
N i=0

60
and all continuous functions f : Sp−l (1) → R.
On the other hand, the arguments of the proof of Lemma 6.3, i.e., that
ψn acts as a permutation on each partition of Sp−l (1) into disjoint balls if
and only if hni = Fp?2 , proves that if n is not a generator of Fp?2 then the
system is not ergodic with respect to the Haar measure. Consequently if ψn
is ergodic then hni = Fp?2 , so that the system is minimal by Theorem 6.2 and
hence even uniquely ergodic by Theorem 6.3. Since unique ergodicity implies
ergodicity one has the following.
Theorem 6.4. The monomial dynamical system ψn : x 7→ xn on Sp−l (1)
is ergodic with respect to the Haar measure if and only if it is uniquely ergodic.
Even if the monomial dynamical system ψn : x 7→ xn on Sp−l (1) is ergodic
it can never be mixing, especially not weak mixing. This can be seen from the
property that an abstract dynamical system is weak mixing if and only if the
product of such two systems is ergodic. If we choose a function f on Sp−l (1)
and define a function F on Sp−l (1) × Sp−l (1) by F (x, y) := f (log x/ log y)
(which is well defined since log does not vanish on Sp−l (1)), we obtain a
non-constant function satisfying F (ψn (x), ψn (y)) = F (x, y). This shows that
ψn × ψn is not ergodic, and hence ψn is not weak-mixing with respect to any
invariant measure, in particular the restriction of the Haar measure.
Let us consider the ergodicity of a perturbed system

ψq = xn + q(x), (45)

for some polynomial q such that q(x) equals to 0 mod pl+1 , (|q(x)|p < p−(l+1) ).
This condition is necessary in order to guarantee that the sphere Sp−l (1) is
invariant. For such a system to be ergodic it is necessary that n is a generator
of Fp?2 . This follows from the property that for each x = 1 + al pl + ... on
Sp−l (1) (so that al 6= 0) the condition on q gives

ψqN (x) equals 1 + nN al mod pl+1 . (46)

Now ψq acts as a permutation on the p − 1 balls of radius p−(l+1) if and only

if hni = Fp?2 . Consequently a perturbation (45) cannot make a non-ergodic
system ergodic.

18 Newton’s Method (Hensel’s Lemma)

Here we present a p-adic analogue of the Newton procedure to find the roots
of polynomial equations, see, e.g., [40], [107], [111], [163]:

61
 Theorem 7.1. Let F (x), x ∈ Zp , be a polynomial with coefficients
Fi ∈ Zp . Let there exist γ ∈ Zp such that

F (γ) = 0 (mod p2δ+1 ) and F0 (γ) = 0 (mod pδ ), F0 (γ) 6= 0 (mod pδ+1 ),

where δ is a natural number. Then there exists a p-adic integer α such that

F (α) = 0 and α = γ (mod pδ+1 ).

Corollary 7.1. (Hensel Lemma). Let p(x) be a polynomial with p-adic

integer coefficients and let there exist γ ∈ Zp such that:

F (γ) = 0 (mod p), F0 (γ) 6= 0 (mod p).

Then there exists α ∈ Zp such that

F (α) = 0 and α = γ (mod p).

19 Computer Calculations for Fuzzy Cycles

The following results were obtained with the aid of the complex of p-adic
programs, p-ADIC, which was created by De Smedt, see, e.g., [61], using the
standard software package MATHEMATICA.
Example 8.1. Consider the function ψ3 (x) = x3 in Q5 . Then we found
among others the following fuzzy cycles. Cycles of length 2:

U 1 (2) − U 1 (3); U 1 (7) − U 1 (18); U 1 (57) − U 1 (68).

5 5 25 25 125 125

62
Cycles of length 4 :
U 1 (6) − U 1 (16) − U 1 (21) − U 1 (11);
25 25 25 25

U 1 (2) − U 1 (8) − U 1 (12) − U 1 (3);

25 25 25 25

U 1 (22) − U 1 (23) − U 1 (17) − U 1 (13);

25 25 25 25

U 1 (9) − U 1 (4) − U 1 (14) − U 1 (19);

25 25 25 25

U 1 (7) − U 1 (93) − U 1 (107) − U 1 (43);

125 125 125 125

U 1 (26) − U 1 (76) − U 1 (101) − U 1 (51);

125 125 125 125

U 1 (18) − U 1 (82) − U 1 (112) − U 1 (32);

125 125 125 125

U 1 (24) − U 1 (74) − U 1 (99) − U 1 (49).

125 125 125 125

Cycles of length 20 :
U 1 (6) − U 1 (91) − U 1 (71) − U 1 (36) − U 1 (31);
125 125 125 125 125

−U 1 (41) − U 1 (46) − U 1 (86) − U 1 (56) − U 1 (116);

125 125 125 125 125

−U 1 (21) − U 1 (11) − U 1 (81) − U 1 (66);

125 125 125 125

−U 1 (121) − U 1 (61) − U 1 (106);

125 125 125

−U 1 (16) − U 1 (96) − U 1 (111).

125 125 125

One of the problems which arise in our computer investigations of p-adic

dynamical systems is that we cannot propose a reasonable way of creating
p-adic pictures which can illustrate our numerical results. However, this is
a general problem of the p-adic framework because the human brain can
understand only pictures in real space.
[Random Processing of Information] Random Processing of Information
We have merely discussed the deterministic model of the process of think-
ing. However, the real process of thinking is strongly disturbed by various
noises. There are two kinds of noise: physiological noise and information
noise. Physiological noise is induced by simultaneous work of a huge number
of neurons (physical and chemical processes in the brain), see, for example,

63
[70]. Information noise is induced by external information as well by the
simultaneous functioning of a huge number of dynamical thinking systems,
fs , gα , φβ , ... in brain. For example, I-states generated by the processor gα
can disturb the work of the processor fs . Moreover, combinations of physio-
logical and information noises in brain may exhibit new features. We have
to study iterations of information which are disturbed by some noise ω (this
parameter ω describes all possible noises in brain). In particular, parameters
s of thinking dynamical processors have to be considered as random variables
of noise ω, i.e., s = s(ω). Instead of a deterministic dynamical system, we
have to use random dynamical system, RDS; see [15] on the general theory
of RDS.
We study the problem of stability of the process of thinking (in our p-adic
dynamical model) with respect to noise (physical, physiological, psycholog-
ical). Our main mathematical result is that for a large class of dynamical
systems in p-adic spaces the process of noisy thinking still produces only
deterministic ideas. This result is in the agreement with our conscious expe-
rience: despite the presence of noise, we always produce deterministic ideas
(of course, if noise is not so strong that it totally disturbs the process of
thinking). From the mathematical viewpoint this fact is quite surprising.
The existence of random solutions is the standard feature of random dynam-
ical systems in Euclidean spaces (see, [15]). We may speculate that in the
process of evolution cognitive systems have found that the p-adic metric on
the space of ideas can, in particular, provide the stability of the process of
thinking with respect to noisy disturbances.
Cognitive implications of our mathematical model are presented in sec-
tion 3. One of the main (and unexpected) consequences is that noise plays a
positive role in the process of thinking. Some combinations of different noise
processes can produce ideas which might not appear as results of determin-
istic thinking processes. We may even speculate that brain contains special
random generators which produce random switching of functioning of think-
ing processors, cf. chapter 2. On the other hand, some special choices of
noise processes s(ω) (which are, of course, related to the concrete cognitive
system) may completely disturb the process of thinking despite the low in-
tensity of s(ω). Noise which is destructive for one brain may have positive
effects for other brains. Such effects depend on the structure of I-space XI
and the dynamical systems fs of the concrete brain. In this framework we
discuss such delicate features of human behaviour as creativeness, stress and
depression.
The theory of p-adic RDS which is used in this chapter was developed in
the collaboration with Prof. M. Gundlach, Bremen Institute for RDS, and
our PhD students, [68], [131].

64
20 Random Dynamical Systems
Random dynamical systems (RDS) describe time evolutions in the presence
of noise. The latter is modeled by a measure-preserving transformation θ on
a probability space (Ω, F, P). For technical reasons one usually assumes that
θ is invertible and ergodic. The dynamics of the RDS takes place on a state
space X, which here we assume to be a compact topological space equipped
with the corresponding Borel σ-algebra of X. In discrete time an RDS φ on
X is then given by products of random continuous mappings φ(ω), ω ∈ Ω.
These are chosen in a stationary fashion according to the noise model, i.e.
the time evolution is given for n ∈ N by

x 7→ φ(n, ω)x = φ(θn−1 ω) ◦ . . . ◦ φ(ω)x

such that (ω, x) 7→ φ(ω)x is measurable. φ defines a measurable cocycle:

φ(n + m, ω) = φ(n, θm ω) ◦ φ(m, ω) for all ω ∈ Ω, n, m ∈ N.

This is the adaptation of the flow property of dynamical systems to the

random situation. For the description of motion the simplest invariant sets,
in particular if they are attracting, are of major interest. In the deterministic
case these are given by fixed or periodic points. Those play a minor role in
random dynamical systems. Note for example that a point x can only be
a fixed point of a random dynamical system φ if it is a fixed point for all
random maps φ(ω). This situation does not occur in general, but we will soon
meet it in p-adic RDS. The situation for periodic points is even worse: a finite
number of iterates would determine the whole orbit which usually contradicts
the random nature of the system. In return there are other notions which
gain importance for RDS, namely stationary solutions, which can be seen
as random analogues of fixed points. These are given by random variables
x : Ω → X such that φ(ω)x(ω) = x(θω) for all ω ∈ Ω. Another way to look
at this phenomenon is to consider the Dirac measures δx(ω) and to integrate
them with respect to P in order to obtain a measure which is invariant for
the RDS and hence a very natural object in this theory. Many phenomena in
elementary stochastic dynamics can be better represented by such invariant
measures than by invariant or stationary subsets of the state space, which in
fact correspond to the supports of the measures. The main advantage is that
the measures reflect the dynamics, while the invariant sets are static objects.
We will encounter this later on in the study of p-adic RDS.
The invariant sets A for RDS φ are in general random, i.e. they will
depend on chance in the sense that they are measurable functions A(ω)
satisfying φ(ω)A(ω) = A(θω) and attracting random neighborhoods U (ω) in

65
the sense that
lim ρ(φ(n, θ−n ω)U (θ−n ω), A(ω)) = 0
n→∞

for P-almost all ω ∈ Ω, a situation which henceforth is abbreviated by P-a.s.

Here we have used the usual Hausdorff pseudometric (see chapter 2) given
by
ρ(D, A) = sup inf |x − y|p .
x∈D y∈A

Though it looks quite unusual, this form of convergence, also known as pull-
back convergence, can be nicely investigated. It yields the convergence in
probability forward in time, i.e., for all  > 0 and all random compact sets
C(ω) in some basin of attraction

lim P{ω : ρ(φ(n, ω)C(ω), A(θn ω)) < } = 1.

n→∞

We need not care about these differences, as we will obtain much stronger
convergence results, namely P-a.s. convergence forward in time. We will
dispense with the rigorous introduction of the general attractor notion, in
particular as in our studies of p-adic RDS we will be confronted only with
the case of attractors which are able to attract non-random neighborhoods.
We will study a p-adic RDS which is a stochastic generalization of the
deterministic dynamical system

xn+1 = ψs (xn ), where ψs (x) = xs , s = 2, 3, ..., x ∈ X, (47)

where X is a subset of Qp . We start with the following technical result.In

principle, we can obtain it by using Lemma 3.1, chapter 3. However, we prefer
the direct derivation to demonstrate again working with p-adic numbers.
Lemma 1.1. Let γ ∈ S1 (0) and u ∈ Zp , |u|p ≤ p1 . Then

|(γ + u)n − γ n |p ≤ |n|p |u|p

for every n ∈ N, where equality holds for p > 2.

Proof. Let n = mpd , where p does not divide m. Define g : x 7→ xm , and
h : x 7→ xp . Then
m  
m m
X m m−k k
|g(γ + u) − g(γ)|p = |(γ + u) − γ |p = | γ u − γ m |p
k=0
k
= |muγ m−1 − o(u2 )|p = |m|p |u|p |γ m−1 |p = |u|p

(where o(z) means terms of p-order smaller than or equal to the order of z,
which here is simply all the rest of the binomial expansion). Thus the map g

66
is an isometry. Set v := g(γ + u) − g(γ),

and y := g(γ). The prime number p
divides all the binomial coefficients kp for 1 < k < p, thus we have for p > 2

|h(y + v) − h(y)|p = |pvy p−1 + o(pv 2 )|p = |p|p |v|p = |p|p |u|p ,

and for p = 2 we have

|h(y + v) − h(y)|p = |pvy p−1 + o(v 2 )|p ≤ |p|p |v|p = |p|p |u|p .

Thus, d iterations of h give

|(γ + u)n − γ n |p = |hd (g(γ + u)) − hd (g(γ))|p ≤ |n|p |u|p ,

where equality holds for p > 2.

As we have already seen (Theorem 5.1, chapter 1), the roots of unity in
Qp play the important role in the investigation of the dynamics of monomial
maps in the p-adic integers. Note that xp−1 = 1 has p − 1 simple solutions.
This is the set of the (p−1)th roots of unity – Γ(p−1) . There exists a primitive
root ξ such that
Γ(p−1) = {1, ξ, ξ 2 , . . . , ξ p−2 }.
For any natural number k, consider the fixed points of the monomial map
x 7→ xk . They are given by xk = x, and so besides of the point x = 0 we
have the solutions of the equation xk−1 = 1. This is the set Γ(k−1) . Note that

Γ(k−1) = {1, ξ m , ξ 2m , . . .} ⊆ Γ(p−1) ,

p−1
with m = (p−1,k−1) , where for a, b ∈ N, (a, b) denotes the greatest common
divisor of a and b, see Proposition 5.1, chapter 1. Given two maps ψk : x 7→ xk
and ψl : x 7→ xl , ψl maps Γ(k−1) into itself, and we have

ψl [Γ(k−1) ] = Γ(u−1) ⊆ Γ(k−1)

k−1
with u = (k−1,l) + 1. So the map ψl acts as a permutation on Γ(k−1) if and
only if the greatest common divisor of k − 1 and l is 1, that is (k − 1, l) = 1.
The points in Γ(k−1) are attracting if and only if p divides k (see Theorem
5.1 of chapter 1).
Now we construct an RDS corresponding to (47) with randomly changed
parameter s. Let s(ω) be a discrete random variable that yields values sj
with probabilities qj > 0, j = 1, ..., m, where sj ∈ N, sj 6= si for j 6= i. We
set φ(ω)x = xs(ω) , x ∈ Qp . This random map generates an RDS
n−1
Y
Sn (ω)
φ(n, ω)x = x with Sn (ω) = s(θi ω), n ≥ 1, x ∈ X, (48)
i=0

67
where X is a subset of Qp . Let us introduce the set (forward orbit)
k1
···skmm
Os (η) = {a ∈ Γ(p−1) : a = η s1 , kj = 0, 1, ...}

of points which can be reached from η evolving due to the RDS, and the set
(backward orbit)
k1
···skmm
Os− (η) = {γ ∈ Γ(p−1) : γ s1 = η for some kj = 0, 1, ...}.

of points which can reach η evolving under the RDS. Let us remark that
this is not a property concerning the asymptotic behaviour, as the points in
Os− (η) indeed reach η in finite time. As usual, due to the invertibility of θ
we can consider φ(n, θ−n ω) = xS−n (ω) , where

S−n (ω) = s(θ−1 ω) · · · s(θ−n ω).

Because of commutativity we have the presentation

m
k (ω)
Y
Sn (ω) = sj j,n
j=1
Pm
for some 0 ≤ kj,n ≤ n with j=1 kj,n = n. From Poincaré’s Recurrence
Theorem we know that for every j ∈ {1, . . . , m}

kj,n (ω) → ∞, for n → ∞ P-a.s. (49)

We are only interested in attractors of random dynamical systems. Therefore,

everywhere below we only consider the case when at least one of sj , j =
1, 2, ..., m, is divisible by p.
As for deterministic systems (47), it is easy to prove that a0 = 0 and a∞ =
∞ are attractors of the RDS (48) with the basins A(0) = U1/p (0) and A(∞) =
Qp \ U1 (0) respectively. These attractors are deterministic in the sense
that supx∈A(0) |φ(n, θ−n ω)x|p → 0, n → ∞, and supx∈A(∞) |φ(n, θ−n ω)x|p →
∞, n → ∞, P-a.s. Hence, as in the deterministic case, we have to study
the behaviour of (48) only on the unit sphere X = S1 (0). We will show
that the RDS has also only deterministic invariant sets, but the dynamics is
stochastic. We remark that

sup |φ(n, θ−n ω)x|p = ρ(φ(n, θ−n ω)A(0), {0}),

x∈A(0)

where ρ is the Hausdorff distance on the collection Sub(Qp ) of all subsets of

Qp .

68
 A set A ⊂ S1 (0) is said to be s-invariant, if ψsj (A) = A for all j =
1, . . . , m.
Define Is := ψsp−11
◦ . . . ◦ ψsp−1
m
(Γ(p−1) ). Is is a cyclic subgroup of order q
(p−1)
of Γ , where q is the greatest divisor of p − 1 with (q, sj ) = 1 for all j, i.e.
Is = Γ(q) . So this is an s-invariant set, since every ψsj is a permutation on
Is .
Example 1.1. Let p = 61, s1 = 61, s2 = 2. Then p − 1 = 60 = 22 · 3 · 5,
and I(61,2) = Γ(14) = {1, ξ 4 , . . . , ξ 56 } for ξ being a primitive 60th root of unity.
If we now add some exponent s3 with (s3 , |I(61,2) |) = 1 (where | · | denotes
the order of the group), then I(61,2,s3 ) = I(61,2) . If we add, e.g., some s3 with
(s3 , |I(61,2) |) = 5, the set I(61,2,s3 ) has order 3 and is equal to {1, ξ 20 , ξ 40 }.
(Some other examples will be considered in the next section).
Theorem 1.1. The set Is is the attractor for RDS (48) on the sphere
X = S1 (0).
Proof. By the above, φ(n, ω)(Is ) = Is , and Os− (Is ) = Γ(p−1) by defini-
tion. Thus it remains to show

lim ρ(φ(n, θ−n ω)X, Is ) = 0 P − a.s.

n→∞

To this end, for every x ∈ S1 (0) set x := γ + u for γ ∈ Γ(p−1) and some u
with |u|p ≤ p1 . Note that γ S−n ∈ Is with probability 1 after a finite number of
steps, and thus, for n sufficiently large, by the Poincaré Recurrence Theorem

ρ(φ(n, θ−n ω)X, Is ) = sup inf |φ(n, θ−n ω)x − z|p

x∈S1 (0) z∈Is

= sup inf |xS−n (ω) − z|p

x∈S1 (0) z∈Is

= sup inf |(γ + u)S−n (ω) − γ S−n (ω) |p

(p−1)
|u|p ≤ p1 γ∈Γ

≤ sup |S−n (ω)|p |u|p → 0 P − a.s.

|u|p ≤ p1

The last inequality holds by Lemma 1.1.

Note that Theorem 1.1 does not make any assertions on the dynamics
apart from where they are concentrated. It just describes a static pattern.
A more complete picture of the attractors of the RDS can be drawn, if we
interpret A as support of an invariant measure µ which can also be obtained
as an attractor of measures. The description of the stochasticity of dynam-
ics can easily be obtained by the upcoming proposition and the invariant
measures should be in accordance with this description. Let us mention the
special case of noise being modeled by a Bernoulli process. This process can

69
be characterized by the following property of the noise flow: the random
variables θ(ω), ..., θn (ω), ... are identically and independently distributed (see
section 3 for the description of such a process).
Proposition 1.1. Let θ be a Bernoulli process and let A be an invariant
subset for RDS (48). Then the dynamics on A is Markovian with homoge-
neous transition probability P (a, b) for the transition from a to b in one step
given by
P (a, b) = P{ω ∈ Ω : φ(1, ω) = ψs , ψs (a) = b}.

Proof. This result is well-known and can be found in a more general

form in [15]. Here we present a simple proof for our situation. From the
presentation of the RDS as products of random maps which are independent,
it is clear that the conditional probability P (ak , nk |ak−1 , nk−1 , . . . , a0 , n0 ) for
a state ak at some integer time nk knowing the previous states ak−1 , . . . , a0
at integer times nk−1 > . . . > n0 ≥ 0 is given by

P (ak , nk |ak−1 , nk−1 , . . . , a0 , n0 )

= P{ω ∈ Ω : φ(nk − nk−1 , θnk−1 ω)ak−1 = ak }
= P (ak , nk |ak−1 , nk−1 ) = P (ak , 1|ak− , 0),

since θ is measure preserving.

Invariant sets. The set Is usually splits into smaller invariant subsets,
in the sense Is = I1 ∪ . . . ∪ In , Ii ∩ Ij = ∅, i 6= j, and ψsj (Ik ) = Ik for all k
and j. {1} is always an invariant set. The basin of attraction of a set Ik is
the set
Os = ∪η∈Ik Os− (η).
Denote the order of the attractor Is by q. Is has itself a primitive root ζ
p−1
which generates it (set ζ := ξ q ). Now consider ψsi -invariant subsets; they
are given by the orbits Osi (ζ a ), a ∈ {1, . . . , q − 1}. What do they look like?
This is the same as asking for the set {a · ski mod q, k ∈ N}. This problem can
usually only be solved numerically. We can give a qualitative answer of which
lengths of invariant sets can be expected. Let da be the number of elements
in the above orbit. Let q = pn1 1 · . . . · pnuu be the unique factorization of q into
primes. Since (si , q) = 1, da is the order modulo q/(a, q) of si , and for this,
it divides the number qa of multiplicatively invertible elements in the ring
mi −1
Z/q/(a, q)Z. Let q/(a, q) = p1m1 ·. . .·pm u
u . Then qa = Πi=1 pi
u
(pi −1) by well-
known number-theoretic considerations. So we know that the lengths of all
orbits divide the numbers qa , a ≤ q − 1. If for example q is prime, (q, a) = 1
for all a, and hence the lengths of the orbits divide q − 1. Examples are
contained in the next section.

70
 The invariant sets of the RDS (48) are appropriate unions of those ψsi -
invariant sets.
It is interesting that the attractor is determined by the greatest common
divisors of the exponents sj and the number (p − 1), and that the invariant
sets and the basins of attraction are determined by the orders modulo q of
sj . So for a given RDS with (s1 , . . . , sm ) we can add the numbers t ∈ N
with t ≡ sj mod (p − 1) for some j to the parameter set (or exchange the
corresponding parameters). This does not change anything of the structure
of invariant sets, but it may change the dynamical behaviour outside. Hence
we can extend the class of RDS by considering infinite sets of parameters,
i.e., s(ω) = sj , sj 6= si for i 6= j, j = 1, 2, ..., with probabilities qj > 0 which
sum up to 1, and at least one of sj is divisible by p. We set s = (sj )j∈N and
∞
k1 k
···sj j ···
X
Os (η) = {a ∈ Γ(p−1) : a = η s1 , kj = 0, 1, ..., kj < ∞};
j=1

∞
k k
s1 1 ···sj j ···
X
Os− (η) = {γ ∈ Γ(p−1)
:γ = η for some kj = 0, 1, ..., kj < ∞}.
j=1

A set A ⊂ S1 (0) is said to be s-invariant, if ψsj (A) = A for all j. By

using Poincaré’s Recurrence Theorem for the random variable s(ω) (having
an infinite number of values) and repeating the proof of Theorem 1.1 we find
that this theorem is also valid for the RDS generated by s(ω).
We discussed random dynamics of I-states. These results can be easily
reformulated for dynamics of ideas. Each I-state a ∈ Γ(p−1) can be identi-
fied with the association of the first order Aa0 ≡ U1/p (a0 ), a = (a0 , a1 , ...).
Therefore each invariant set-attractor I ⊂ Is can be identified with the idea
JI = {Aa0 }a∈I . The basin of attraction A(I) gives the basin of attraction
A(JI ) which is equal to {Aa0 }a∈Os . In fact, we proved that random itera-
tions of the idea (set of associations) A(JI ) converge to the idea-attractor
JI . Therefore we have
Theorem 1.2. RDS (48) has only deterministic attractors in the space
of ideas XID .
Therefore the presence of randomness could not disturb the production of
deterministic ideas. Despite random fluctuations in the process of thinking,
brain still produces the same idea-attractor starting with the fixed initial
idea. Of course, as in all probability one statements, we do not have 100%
determinism.

71
21 Longterm Behaviour, Dynamics on the At-
tractor, Examples
In this section we consider the long-term behavior of some examples of p-adic
RDS which have an attractor due to Theorem 1.1. Fix a prime number p,
denote by ξ the primitive root of unity of degree p − 1. According to the
above, we only need to consider parameters sj ≤ p. We also leave aside the
parameters s = 1 (corresponding to the identity) and s = p − 1 (for which
the attractor is {1}). Now let

s : Ω → {s1 , . . . , sm }

be a random
P variable with a distribution given by (q1 , . . . , qm ), such that
qi > 0, i qi = 1. The RDS φ is given by
 S (ω)
 x n , n ≥ 1,
φ(n, ω)x = x, n = 0,
 S−n (ω)
x , n ≤ −1.
As random selection mechanism we choose an m sided dice corresponding to
the probability distribution (q1 , . . . , qm ) . The dice is thrown independently
in each time step. This type of random influence can be modeled by a so
called Bernoulli shift, which is a measure-preserving transformation θ on the
space of all two-sided sequences consisting of m symbols.
Because of Theorem 1.1 and Proposition 1.1 we can restrict our consider-
ations to the motion of φ on the attractor Is where the dynamical behaviour
can be described by a (possibly inhomogeneous) Markov chain. According to
the choice of the random selection mechanism in our examples the resulting
Markov chain is homogeneous, i.e. the transition probability does only de-
pend on the current state and is independent of time and chance. Now, the
long-term behavior of this Markov chain is determined by a stationary distri-
bution. Such a stationary distribution always exists due to the fact that the
transition matrix of the Markov chain has the number 1 as an eigenvalue. It
might be nonunique if the Markov chain is not irreducible, where irreducibil-
ity means that there is a positive probability for each state to reach any other
state. It is easy to see that the Markov chain given by φ on Is cannot be
irreducible, since ξ 0 = 1 is always a fixed point which is never left if it is hit
once.
If a fixed point is reached, the dynamics of φ can be considered as a
trivial Markov chainMarkov chain on one state. If there are some φ-invariant
subsets of Is on which φ acts as a nontrivial Markov chain, we can separate
the attractor into components on which the dynamical behaviour of φ is the

72
one of an irreducible Markov chain. In this case the stationary distribution on
such components is unique and determines the motion of φ, but the selection
of the finally attained component depends on the initial conditions.
Let us look at the RDS φ with p = 29 and s1 = 29, s2 = 2, s3 = 3. Since
p − 1 = 28 = 22 · 7 we obtain the attractor as I(29,2,3) = {1, ξ 4 , ξ 8 , . . . , ξ 24 }
consisting of q = 7 elements where ξ is the primitive 28th root of unity. The
order of 2 modulo 7 is 3, and the order of 3 is 6. Thus we know that in I(29,2,3)
there are 2 ψ2 -invariant sets and 1 ψ3 -invariant set beside {1}. This means
I(29,2,3) splits into the two (29, 2, 3)-invariant sets {1} and {1, ξ 4 , ξ 8 , . . . , ξ 24 }.
If we look at the dynamics of ψ2 (x) = x2 on the attractor we see the fixed
point 1 with domain of attraction {ξ 7 , ξ 14 , ξ 21 } and two invariant subsets
{ξ 4 , ξ 8 , ξ 16 } and {ξ 12 , ξ 24 , ξ 20 } with domains of attraction

{ξ, ξ 2 , ξ 9 , ξ 18 , ξ 11 , ξ 22 , ξ 15 , ξ 23 , ξ 25 }
and

{ξ 3 , ξ 6 , ξ 5 , ξ 10 , ξ 13 , ξ 26 , ξ 17 , ξ 19 , ξ 27 }
respectively. Doing the same for ψ3 we obtain the fixed point 1 and a 6-cycle
consisting of J := I(29,2,3) \ {1}. Due to this 6-cycle for ψ3 both invariant
components of ψ2 are merged together such that the attractor of the RDS φ
consists of two components on which the dynamics is given by an irreducible
Markov chain: The set J and the fixed point 1. Thus we have the following
picture of the Markovian dynamics on the attractor I(29,2,3) :
q3
# # q#
q2 - q3
ξ4 ξ8  ξ 12
"! "! q2 "!
I q3
@ @  #
6 @ @ q2 +-q3
q2 @
@
@ q2 ξ0
q2 @ q# "!
@
3 ?
# @ #
@
R
@
q3 - q2
ξ 16 ξ 20  ξ 24
"!
i "! "!
q3

Figure 3: The Markov chain given by φ on I(29,2,3) , (q1 is omitted).

Note that (by symmetry arguments) the transition matrix for the Markov
chain on J must be bistochastic (i.e. row as well as column sums are 1).
Hence the left eigenvectors to the eigenvalue 1 have identical coordinates and

73
yield the equidistribution as the unique stationary measure for the Markov
chain. This assigns probability 16 to each element of J independent of the
probability distribution (q1 , q2 , q3 ) of our selection mechanism. Whether the
motion finally reaches the fixed point or whether it remains in J depends on
the initial conditions of the RDS. Thus we have determined all the invariant
measures of the RDS φ: First the Dirac measure supported on the fixed point
1, and second the stationary distribution on J . Those are the ergodic in-
variant measures of φ. All other invariant measures are convex combinations
of these two measures.
Let us now return to the example with p = 61 and s1 = 61, s2 = 2.
As we have seen above the attractor I(61,2) consists of 15 elements, where we
observe the unique fixed point 1, one invariant subset consisting of 2 elements
and three subsets consisting of 4 elements each. Again, the ergodic invariant
measures of φ are the unique stationary distributions on these components,
which again are all symmetric. As already discussed the size of the attractor
shrinks to 5 elements if we add s3 = 3 to the RDS φ. The attractor I(61,2,3)
consists of the fixed point 1 and the set {ξ 12 , ξ 24 , ξ 48 , ξ 36 } on which φ acts
as an irreducible Markov chain. Thus the extended φ again has two ergodic
invariant measures, similar to the above example.
In general, these phenomena can be observed if we increase the influence
of noise, i.e., if we allow the random variable s to take more different values.
But, if the set of values of s becomes too large, the whole of S1 (0) is attracted
to the fixed point 1. Summarizing our experimental results, we can say that
stronger noise leads to a decrease of the attractor size as well as the number
of invariant subsets. In particular the fixed point ξ 0 remains as the only
attractor if the noise becomes too strong. On such invariant subsets φ acts
as an irreducible Markov chain. Here a stationary distribution assigns the
same probability to all members of this particular subset. The selection of
the irreducible component depends on the initial conditions and on chance.
The choice of the probabilities qi for the RDS φ only affects the time until
the irreducible component is reached.

22 Consequences for Cognitive Sciences

We studied random dynamics of I-states and the corresponding dynamics of
ideas. Our mathematical investigations have the following cognitive implica-
tions:
1. The presence of noise cannot disturb the clearness of thinking. A
p-adic dynamical cognitive system can never produce ‘random ideas.’
2. Attractors of p-adic RDS can be very complicated configurations of

74
I-states (= ideas). For example, let p = 41, s1 = 11, s2 = 41, then there
are 25 invariant subsets (10 fixed points and 15 sets with 2 points); let p =
41, s1 = 17, s2 = 41, then there are 16 invariant subsets (8 fixed points and
8 sets with 4 points); p = 41, s1 = 13, s2 = 41, then there are 14 invariant
subsets (4 fixed points, 2 sets with 2 points and 8 sets with 4 elements).
3. The presence of randomness in the p-adic cognitive system has im-
portant positive effects for the thinking process. First, fluctuations of the
parameters sj exclude the possibility of transient behaviour, i.e. the stabi-
lization of regimes which do not possess attractors (if qj > 0 at least for one
sj , (p, sj ) 6= 1, then all invariant sets of the RDS are attractors). The second
positive effect is that fluctuations of sj can produce new configurations of
I-states (ideas) which are not available for deterministic dynamical systems.
4. The levels of probabilities qj > 0 do not determine the concrete con-
figuration of I-states (ideas) but only the speed of the thinking process.
5. Our model can be used to explain why individuals may have different
thinking abilities. Some values sj produce very rich configurations of I-
states (ideas). For example, if p = 29 and s1 = 15, s2 = 29, then the
attractor decomposes into 21 invariant sets (attractors). Many other values
of sj produce quite poor configurations of I-states (ideas). For example, if
p = 29 and s1 = 14, s2 = 29, then the attractor is {1}. Thus one cognitive
system might use the noisy control based on the random fluctuations of
‘fruitful parameters.’ Other cognitive system might use the noisy control
based on the random fluctuations of ‘non-fruitful parameters.’ This difference
may on one hand be a consequence of differences in physical structures of
cognitive systems (different physical generators of randomness). On the other
hand, it can also be a consequence of the process of noisy learning. Within
this process cognitive systems record some distributions of chance which they
will use for the thinking process. One cognitive system may be accustomed
to ‘fruitful noises’, other cognitive system to ‘unfruitful noises.’
6. Our simple model already captures essential and realistic features of the
psychological behaviour of human individuals. As we have already remarked,
in the process of noisy learning each cognitive system records a collection of
probability distributions of parameters. This collection determines the range
of ideas which are available for the cognitive system. If a cognitive system
has no ability to change the collection of parameters (in reasonable time), it
will always operate with the same range of ideas. Cognitive systems which
have a larger range of ideas necessarily have a larger ability to change the
distributions of the parameters.
7. On the basis of the p-adic dynamical cognitive model we might try to
explain the roots of some mental diseases. Some values of the parameters sj
imply a strong degeneration of ideas. As we have remarked, if p = 29 and

75
s = 14 then the whole mental space, XI = S1 (0), is attracted by the unique
I-state (idea) a = 1. This situation can be related to the human manic
behaviour.behaviour manic Moreover, such parameters of degeneration are
stable with respect to noise. For example, if p = 29 and s1 = 14 (with
reasonably high probability q1 ), then a noise on any set of parameters sj , j =
2, ..., m, will imply attraction of X = S1 (0) to the same idea a = 1. Thus
it is practically impossible to change the manic behaviour by noisy learning.
Only the consciousness can eliminate the parameter of degeneration. The
value of probability qj > 0 of a parameter of degeneration determines the
‘level of manic behaviour’.
8. We might also explain the so called crises of creativity.crises of cre-
ativity Let τ be a cognitive system which has RDS with a rich configuration
of attractors and let τ be disturbed in such a way that a new RDS has a
poor structure of attractors. This can be done by the appearance of just
one new parameter sj (with qj > 0) in the distribution of noise. In this case
the ‘creativity’ of τ will be completely destroyed. It seems that only the
consciousness can eliminate this parameter and recover the creativity.
9. Our model might explain the origin of stress. Trying to obtain new
configurations of I-states (= ideas), the consciousness uses new random mix-
tures of parameters sj , j = 1, ..., m. It is quite natural to operate with larger
and larger numbers m of parameters (by using rather small probabilities
qj > 0). However, this process which involves new parameters will arrive
at some parameter of degeneration. Since this instant of time, the cognitive
system will work very actively (as a large number of parameters is involved in
the process of random thinking) and rather non fruitful. This situation can
be related to the psychological state of stress. Moreover, our model implies
that in this state the cognitive system will automatically concentrate on a
very restricted set of ideas (or only one idea, Idee fixe). Such a concentra-
tion can be related to some human mental diseases. The only way to change
the situation is to eliminate the parameters of degeneration. However, as a
large number of parameters are involved in the process of random thinking, it
is practically impossible to find the parameters of degeneration consciously.
Thus the cognitive system uses the following simple method: It eliminates
as many parameters as possible from the thinking process (a ‘total relax-
ation’). In this way it can eliminate parameters of degeneration. If stress is
not stopped (i.e. the ‘total relaxation’ did not eliminate some parameters of
degeneration), then the cognitive system repeats the procedure of the ‘total
relaxation’ with respect to another set of parameters.
[Information Quantum Mechanics] Information Quantum Mechanics
Since the creation of quantum mechanics there have been permanent dis-
cussions about the possible connections between quantum and mental phe-

76
nomena. During the last hundred years, a huge number of various proposals
and speculations have been presented. We shall mention just a few of them.
The philosophical system of WhiteheadWhitehead [195]–[197] was the
first attempt to establish a quantum/mental connection. Whitehead tried
to explain a rather unusual statistical behaviour of quantum systems by
speculating that quantum systems are cognitive systems (at least in some
generalized sense), see also ShimonyShimony [175]. An extended discussion
on a quantum/mental connection was induced by attempts to solve the prob-
lem of quantum measurements, see, e.g., [90], [172], [194]. The most extreme
point of view is that physical reality is, in fact, created by acts of obser-
vations. This kind of considerations is especially closely related to the so
called orthodox Copenhagen interpretation of quantum mechanics. By this
interpretation a wave function provides the complete description of an indi-
vidual quantum system. An act of measurement induces collapse of the wave
function. There are various ideas connecting quantum collapse and an act
of thinking, see, e.g., Orlov [148] (quantum logic of consciousness); see also
PenrosePenrose [150]:
“I am speculating that the action of conscious thinking is very much
tied up with the resolving out of alternatives that were previously in linear
superposition.”
In fact, Penrose, as well as many others (see further references), used
the reductionist approach, quantum reductionism.quantum reductionism. By
this kind of reductionism it is impossible to reduce cognitive phenomena to
the neural activity. However, it might occur that we could reduce cognitive
phenomena to activity of quantum systems. Roughly speaking, an act of
thinking is reduced to the collapse of the wave function in quantum gravity.
Our thinking ability is based on collapses of superpositions of two mass states.
The idea of quantum-physical reduction for cognitive processes is quite
popular in the quantum community. We also mention the investigations of H.
Stapp [178] who used the Copenhagen (Heisenberg potentiality) approach to
quantum mechanics. He also used the quantum reductionist approach: ‘Brain
processes involve chemical processes and hence must, in principle, be treated
quantum mechanically.’ We should also mention quantum field reductionist
models, Jibu and Yasue [100] (based on Umezawa [185]) and Vitiello [186].
These quantum field models look more attractive (at least to me). At the
moment there is no idea how to make the great jump from individual gravi-
tational collapses to global acts of cognition. Quantum field models are more
useful in providing such a global structure connecting individual quantum
collapses with global acts of thinking.
However, it seems that reductionism as an general methodology of brain’s

77
study is less and less popular in the cognitive sciences. After the period of
great hopes associated with new possibilities of studying neuron firings, there
is strong disillusionment with the possibility of some physical reduction of
mental processes. This is one of main reasons for the rather strong critical
attitude against quantum models in the cognitive sciences. In the extreme
form this criticism is expressed in the following form:
The only common thing between the quantum and the mental is that we
have no idea of how to understand any of these phenomena.
The other thing that induces prejudice against quantum reduction the-
ories among neurophysiologists is that quantum micro-description contains
many parameters whose magnitudes are far from the magnitudes of the cor-
responding parameters of the brain (e.g., temperature, time scale, and so on).
Of course, it may be that all these parameter problems are just temporary
technical problems. Nevertheless, there are doubts about the possibility of
the direct application of quantum physical theory to cognitive phenomena.
Finally, we discuss the holistic approach to cognitive phenomena based on
the Bohm–Hiley–Pilkkänen theory of active information. By considering the
pilot wave as a kind of information field they presented interesting models of
cognitive processes, see [35]–[37], [92], see also the author’s work [120] and
chapter 6. Consciousness–information models were also developed in the
books of M. Lockwood [132], and J. A. Barrett [24] (who use a many-minds
version of the many-worlds interpretation of quantum mechanics) and the
author’s paper [116] and chapter 6 devoted to quantum information reality.
For the last few years I have tried to split, see [123]–[124], the quantum
formalism into two more or less independent parts:
1) the truly quantum (quanta, Planck constant, discreteness),
2) the Hilbert space probabilistic formalism.
The reader who has read the original papers of Einstein on the foun-
dations of quantum theory would agree that they were merely investiga-
tions on discreteness (quantization) of energy. The quantum probabilistic
(Hilbert space) formalism was developed later (BornBorn, HeisenbergHeisen-
berg, DiracDirac). It was created to describe the statistics of elementary
particles (quanta)— Hilbert space probabilistic calculus is always related to
quantaquanta, quantum probability, and so on. Careful analysis [123]–[124]
demonstrated that Hilbert space probabilistic calculus Hilbert space proba-
bilistic calculus (Born, Heisenberg, Dirac) is a purely mathematical formalism
which gives the possibility of working with context dependent probabilities,
i.e., probabilities depending on complexes of physical conditions (contexts)
related to concrete measurements. In this framework the interference of alter-
natives is the interference of two different contexts, see [123]–[124]. Therefore
we could apply the Hilbert space probabilistic formalism, a quantum-like for-

78
malism, not only to the description of statistical micro-phenomena, but also
to various phenomena outside the micro world. One of such possibilities is
to apply a quantum-like formalism to the description statistical experiments
with cognitive systems. Such an approach has no (at least direct) relation
to reductionist quantum models. We are not interested in the statistical be-
haviour of micro-systems forming a macro-system, the brain. Therefore this
approach does not induce such a problem as the transition from micro to
macro (temperature and so on). We just use Hilbert space probabilistic for-
malism to describe information (in particular, cognitive) measurements. As
in the ordinary quantum formalism, information observables, I-observables,
are realized as symmetric operators in the Hilbert space of square integrable
functions φ(q) depending on the I-state q of a cognitive system. By using
the Hilbert space scalar product we calculate averages of I-observables. Of
course, this cognitive model is a purely statistical one. It could not provide a
description of individual thought trajectories, cf., chapter 6 — the Bohmian
model for informational quantum mechanics.
We underline that the main reason for using a quantum-like formalism
to describe the statistics of measurements over cognitive systems is that
cognitive systems (as well as quantum systems) are very sensitive to changes
of context of an experiment — a complex of physical and cognitive conditions,
compare with Heisenberg views on ordinary quantum measurements.
As we have already discussed in the details in chapter 1, one of the funda-
mental problems in the foundations of information processing (in particular,
cognitive models) is the choice of a mathematical model for the configuration
I-space. In quantum-like I-models this is the problem of a mathematical rep-
resentation of the space of classical I-states, namely, the space on that wave
function is defined. We have already underlined that the Euclidean physical
space (in which the physical brain is located) does not look attractive as a
model of I-space. Instead of this conventional model of space we develop a
cognitive quantum-like formalism on the space of information strings which
could be performed by chains of hierarchically ordered neurons, see chapter
2. Such a configuration space is geometrically represented by a hierarchic
p-adic tree. The p-adic I-space was used in chapters 1,2, 4 for the mathe-
matical modelling of classical I-processes. In this chapter we shall develop a
kind of quantum formalism on the p-adic configuration space. In this quan-
tum model a wave function is defined on a p-adic tree, but take values in the
ordinary set of complex numbers,
φ : Q → C, where Q = Zp or Qp .
In fact, the mathematical formalism of such a model of p-adic quantum me-
chanics (the Vladimirov–Volovich Model) was already well developed, see

79
[187]. We apply this formalism to information (in particular, cognitive) phe-
nomena.
In the ordinary quantum mechanics we could go beyond the statistical ap-
plication of the quantum formalism. One of the most attractive possibilities
is to use the Bohmian pilot wave formalism. As we have already remarked,
the idea to use Bohmian mechanics in cognitive sciences was already well
discussed (Bohm–Hiley–Pilkkänen [92] and author [120]). It is rather sur-
prising that it seems to be impossible to create a variant of the pilot wave
extension of quantum-like I-formalism presented in this chapter (but cf. with
chapter 6). Formally we can introduce a quantum-like I-potential and force.
However, there is no possibility of deriving the equation of motion (a kind
of Newton’s equation) that would describe trajectories of individual I-states
(describe flows of mind). In our formalism this is a consequence of the math-
ematical structure of the model. However, it may be that there are some
deep cognitive features behind this mathematical result, cf., chapter 6.

23 Quantum-Like Formalism for a One-Layer

Brain
1.1. Mental configuration space for a one-layer brain.
We consider the simplest hierarchical ‘brain’ consisting of just one hier-
archic chain of neurons: N = (n0 , n1 , ..., nN , ...), see chapter 2. In a mathe-
matical model it is convenient to consider an infinite chain. In the simplest
model each neuron can represent only one of two states: αj = 1 (firing) and
αj = 0 (not). In more complex models nj can perform p different levels:
αj = 0, 1, ..., p − 1 (for example, frequencies of firing). It is assumed that
neurons in this layer are hierarchically ordered: n0 is the most important (ig-
niting), n1 is less important and so on. The N is able to produce information
strings of the form:

x = (x0 , x1 , ..., xN , ...), xj = 0, 1, ..., p − 1.

We choose the space Q = Zp as configuration I-space. Points q ∈ Q are

called classical-like I-states (or simply I-states) or I-positions.

80
 1.2. Hilbert space probabilistic formalism.
We consider the space of square integrable functions L2 (Zp , dx) :
Z
2
φ : Zp → C, kφk = |φ(x)|2 dx < ∞.
Zp

Here dx is the Haar measure on the compact additive group Zp , see chapter
3.
The space H = L2 (Q, dx) is chosen as the space of quantum-like I-states.
These states are represented by normalized vectors φ ∈ H : kφk = 1. The H
is a complex Hilbert space with scalar product
Z
(φ, ψ) = φ(x)ψ(x)dx. (50)
Q

I-observables (cognitive observablescognitive observables) are realized as self-

adjoint operators A : H → H. As in the ordinary quantum formalism, by
fixing a quantum-like state φ ∈ H in general we do not fix the concrete value
A = λ of the I-observable A. It is only possible to find the average of A in
the state φ : Z
hAiφ = A(φ)(x)φ(x) dx. (51)
Q

However, if φ ∈ H is an eigenfunction of A corresponding to the eigenvalue

λ, i.e., Aφ = λφ, then we can be sure that we shall obtain the value A = λ
with probability 1.
The concrete representations of I-observables by self-adjoint operators is
a very important and nontrivial problem. This problem could not be solved
by trivial generalization of ordinary quantum formalism. We start with the
surprising remark that it seems to be impossible to define the observable of
the I-position q. Formally the difficulty is purely mathematical: we can not
multiply a p-adic number q ∈ Q by a complex number φ(q). Therefore the
standard Shrödinger definition of the position operator can not be gener-
alized to the I-case. Of course, we could try to find some mathematically
tricky definitions of the I-position operator. However, it might be that this
mathematical difficulty is an evidence of some important feature of cogni-
tive systems. It might be that even in principle it is impossible to measure
I-states q of the brain. In particular, we could not prepare the brain in the
fixed I-state — there are no I-state eigenfunctions.
We can only find the probability that an I-state q belongs to some (mea-
surable) subset O of the I-space Q :
Z
P (q ∈ O) = |φ(x)|2 dx.
O

81
 Example 1.1. Let us consider the quantum like state φ ≡ 1 (the uniform
probability distribution of I-states). Then P (q ∈ Ur (a)) = r. Thus (as it
could be expected) the probability to find this cognitive system in the I-
state q belonging to a small ball around any fixed point a is small.

24 Motivation Observable
2.1. Multi-layer hierarchical brain.
To consider nontrivial examples of I-observables it is convenient to study a
‘brain’ described by a more complex I-space. Such a brain consists of a few hi-
erarchical p-adic trees. We consider a layer of neurons N = (..., nk , ..., n0 , ..., nl ,
...) which goes in both directions (in the mathematical model it is infinite
in both directions). Each neuron nj , j = 0, ±1, ±2, ..., can be the igniting
neuron for the right hand side hierarchical chain Nj = (nj , ...., nl , ...). The
corresponding I-space Z(j) consists of all information strings
x = (xj , xj+1 , ..., xl , ...), xl = 0, 1, ..., p − 1
(in particular, Zp = Z(0) ). Each space has the structure of the homogeneous
p-adic tree. These spaces are ordered by inclusion: Z(j+1) ⊂ Z(j) . We remark
that the union of all these spaces is, in fact, the set of p-adic numbers:
Qp = ∪∞ (j)
j=−∞ Z .

Geometrically this space is represented as a huge collection of trees ordered by

the inclusion relation. We now choose Q = Qp as a model of a configuration
I-space.
We consider the Hilbert space H = L2 (Q, dx) of square integrable func-
tions φ : Q → C as the space of quantum-like I-states.
2.2. Motivation magnitude observable.
It would be interesting to consider the following quantity (more precisely,
qualia): the motivationmotivation ξ to change the I-state q. Unfortunately,
by the same reasons as for the I-state observable we can not introduce a
motivation observablemotivation observable. However, we can introduce an
observable Mξ that will give the magnitude of a motivation. It is impossi-
ble to prepare a brain with the fixed motivation ξ, but we could prepare a
brain with the fixed amplitude of a motivation (which gives a measure of a
motivation’s strength). Such Mξ must be a kind of derivative with respect
to the I-state (coordinate) q. Such a generalization of a derivative is given
by Vladimirov’s operatorVladimirov’s operator D, see [187], defined with the
aid of the p-adic Fourier transformWe remark that it is impossible to define
the derivative for maps from Qp to R, see [107]..

82
 p-adic Fourier transform:Fourier transform
Z
F (φ)(ξ) = φ(ξ)
e = φ(x)e(ξx)dx, ξ ∈ Qp ,
Qp

where e : Qp → C is a p-adic additive characteradditive character (an ana-

logue of an exponent): e(ξ(x + y)) = e(ξx)e(ξy) for all x and y belonging
to Q. We remark that all p-adic additive characters can be parameterized by
elements ξ ∈ Qp . Thus the dual group, the group of additive characters, for
Qp can be identified with Qp .
Additive characters on Qp have the form e(ξx) = e2πi{ξx} , ξ ∈ Qp . There
{ξx} is the fractional part of the p-adic number z = ξx, i.e., if z = α−m p−m +
... + α−1 p−1 + α0 + α1 p + ... + ... then {z} = α−m p−m + ... + α−1 p−1 .
As usual, F is the unitary operator

F : L2 (Qp , dx) → L2 (Qp , dx).

Vladimirov’s operator of order α > 0 is defined as

Z
α
D (φ)(x) = |ξ|α φ(ξ)e(−ξx)dξ.
e
Q

We remark that Dα Dβ = Dα+β .

The spectral properties of differential operators in L2 (Qp , dx) were inves-
tigated by Vladimirov, Volovich, and Zelenov [187]. Vladimirov also investi-
gated the spectral properties of the pseudo-differential operators:
Z
P (D, x)φ(x) = P (ξ, x)φ(ξ)e(−ξx)dξ
e
Qp

In applications to quantum mechanics Vladimirov and Volovich [187] also

used the integral operators which describe the evolution of quantum systems:
Z
U f (x) = K(y, x)f (y)dy.
Qp

For example, the evolution of the quantum harmonic oscillator is described

by the integral operator with the kernel:
x2 + y 2
  
1 xy
Kt (x, y) = λp (2t) 1/2 exp 2πi − .
|t|p sin t 2 tan t

This kernel depends on the p-adic time parameter t ∈ Urp \ {0}. The function
λp (2t) has a complicated dependence on p.

83
 We define the observable M of the magnitude of motivation as
Mξ = hD. Here h = p1m is some normalization constant. h plays the role
of the Planck constant in ordinary quantum mechanics. At the moment it is
not clear “Can we expect that there exists a kind of universal constant h, the
mental Planck constantmental Planck constant?” I am quite sceptical that
such a universal normalization constant really exists. It is more natural to
suppose that h would depend on a class of cognitive systems under consider-
ation. In fact, by finding h (the level of discretization of motivation ) we find
the basis p of the coding system I still take into account the possibility that
each human individual has its own basis p of coding of information. So it
may be that I am the 2-adic human being and the reader 3-adic or 1997-adic..
To calculate averages of the motivation magnitude operator M for dif-
ferent quantum-like I-states it is natural to use the Fourier transform. By
analogy with ordinary quantum mechanics we could speak of moving from
the position to the momentum representation. Such a terminology is only
formal in information theory. As we have already discussed, there are no
informational position and momentum observables.
Example 2.1. Let a quantum-like state φ be such that its Fourier trans-
e is uniformly distributed over the ball Ur (0), r = 1/pl . Here
form φ(ξ)
Z
l 1
hMξ iφ = p |ξ|p dξ = l−1 .
Ur (0) p (p + 1)

It is important to remark from the beginning that (opposite to the ordi-

nary quantum momentum) the Mξ is a nonlocal operator. It can be repre-
sented as an integral operator, see [187]:

p2 φ(x) − φ(y)
Z
D(φ)(x) = dy .
p + 1 Q |x − y|2p

To find Mξ (φ)(x) at some fixed point x we have to take into account values
of φ at all points of the configuration space.
We remark that Vladimirov’s operator D has a system of (generalized)
eigenfunctions which is similar to the system of free-wave eigenfunctions in
ordinary quantum mechanics. We recall that in ordinary quantum mechanics
the wave φξ (x) = eiξx/h corresponds to the fixed value ξ of momentum. In
the I-framework we have

Mξ e(hξx) = |ξ|p e(hξx).

Here we have used the result [187] De(ξx) = |ξ|p e(ξx).

84
 We remark that in the ordinary quantum formalism the h is placed in the
denominator, ξx/h, and in the p-adic quantum formalism it is placed in the
nominator, hξx. This is a consequence of 1/h being large in R and h being
large in Qp .
The function φξ (x) = e(hξx) is a kind of free information wave,free in-
formation wave I-wave, corresponding to the fixed value ξ of the motivation.
As |φξ (x)| = 1 for all x ∈ Q, the probability of finding a cognitive system
in the I-state x does not depend on x. By analogy with ordinary quantum
mechanics we would like to interpret this mathematical property in the fol-
lowing way: by fixing the magnitude of motivation (strength of will) we can
not localize the I-state. However, we shall soon see that such an analogy
(between material and I-states) could not be used.
The I-wave φξ (x) is not determined uniquely by the observable Mξ . The
main distinguishing feature of p-adic quantum mechanics (discovered by
Vladimirov, [187]) is huge degeneracy of spectrum of the momentum and
energy operators. In particular, besides eigenfunctions φξ (x) the Mξ has an
infinite set of other eigenfunctions corresponding to the eigenvalue λ = |ξ|p (=
pk for some k = 0, ±1, ±2, ...).
Each λ = pk , k = 0, ±1, ±2, ..., corresponds to an infinite series of eigen-
functions (distinct from the free I-wave φξ (x)) belonging to L2 (Q, dx). We
remark that free I-waves φξ (x) are so called generalized eigenfunctions. They
are not square integrable. Thus they do not belong to the space of quantum-
like I-states H = L2 (Q, dx). These eigenfunctions are well localized (concen-
trated in balls) in the configuration space.
This is very natural from the informational point of view. It would be
quite strange if the only quantum-like I-state that has a fixed motivation
magnitude were the state φξ characterized by a totally indefinite distribu-
tion of I-states q. For intuitive reasons there must be quantum-like I-states
characterized by the fixed Mξ = λ that are concentrated on a special class of
I-states (a kind of special cognitive activity).
One of the most important distinguishing features of quantum-like I-
theory is that the motivation magnitude operator Mξ has a discrete spectrum
(except at one point, see further considerations). Hence the magnitude of
the motivation can not change continuously.
There exists only one point of the spectrum of the operator Mξ that is not
its eigenvalue λ = 0. It is the limit point of the eigenvalues λk = pk , k → ∞.
There is no eigenfunction φ0 belonging to the state space H. Thus in our
model the brain could not be (alive, awake?) in the stationary quantum-like
I-state having the motivation of zero magnitude.
Another distinguishing feature is infinite degeneracy of the spectrum.
This purely mathematical result can have important implications for the

85
problem of correspondence between the mental and physical worlds. In fact,
as a result of this huge degeneracy, we can not uniquely determine the I-state
of a cognitive system by fixing a motivation.

25 Neuron Activation Observable

As we have already discussed, we can not introduce the I-state observable
q. However, in the same way as for the motivation we can introduce the
operator of the p-adic magnitude of an I-state:

Mq φ(x) = |q|p φ(x).

Spectral properties of this operator are similar to spectral properties of the

operator Mξ : discreteness and infinite degeneracy of spectrum. Eigenfunc-
tions of Mq (belonging to H = L2 (q, dx)) are localized in p-adic balls–trees.
Therefore:
there exist stationary states of Mq that are characterized by the activation
of a fixed tree of I-states.
Unfortunately, Mq can not be used to fix such a tree (as a consequence
of the infinite degeneracy of spectrum).
The operators of position and motivation magnitudes, Mξ and Mq , do not
commute (like the operators of position and momentum in ordinary quantum
mechanics):
[Mq , Mξ ] = Mq Mξ − Mξ Mq = hJ,
where J 6= 0 is an integral operator [187]. Thus we obtain the information
uncertainty relationinformation uncertainty relation; compare to [111]:
Theorem 3.1. For any quantum-like I-state φ it is impossible to measure
motivation and position magnitudes with an arbitrary precision.
By measuring the motivation magnitudes we change position magnitudes
and vice versa.
This can also be expressed mathematically by using the p-adic Fourier
transform. We denote by Ωr (x) the characteristic function of the ball Ur (0)
(it is equal to 1 on the ball and 0 outside the ball). We have [187], p. 102,
that
e r (ξ) = 1 Ω1/r (ξ).
Ω
r
Thus we obtain the following theorem:
Theorem 3.2. If the state of mind is concentrated on the ball Ur (0), then
motivations are concentrated on the ball U1/r (0).

86
 As in the case of the Mξ -observable the point λ = 0 belongs to the
non-discrete spectrum of the Mq observable. Thus there is no stationary
quantum-like I-state φ corresponding to zero magnitude of q.
To understand better the informational meaning of the Mq observable it
is useful to consider a new I-observable:

A = − logp Mq .

If φ ∈ H is an eigenstate of the Mq corresponding to the eigenvalue λ = |q|p =

1
pk
, then φ also is an eigenstate of A corresponding to the eigenvalue µ = k
and vice versa. Thus the discrete part of the A-spectrum coincides with the
set of integers Z. The A gives the position of the igniting neuron in a layer
of neurons. It is called the neuron activation observable. We note that there
is an interesting relation between the neuron activation observableneuron-
activation observable and entropy. p
√ Let us consider the quantum-like state φ(q) = (p + 1)|q|p Ω1 (q). Here
p + 1 is just the normalization constant. The corresponding probability
distribution P(q) = (p + 1)|q|p on the tree Zp and equals to zero outside this
tree. The entropy of this probability distribution is
Z
EP = − logp P(q)P(q)dq = hAiφ − logp (p + 1).
Zp

26 Complex Cognitive Systems; Evolution

We now consider a cognitive system consisting of n hierarchical layers of
neurons. It can be an individual brain as well as a system of brains. The
mental space of this cognitive system is Q = Qp × · · · Qp (n times). For
each I-coordinate qj , j = 1, 2, ..., n, we introduce the motivation magnitude
operator Mj = hDj , where Dj is Vladimirov’s operator for qj .
We introduce the P kinetic I-energy (the free energy of motivations) as
H = h ∆, where ∆ = nj=1 Dj2 is the VladimirovianVladimirovian (a p-adic
2

analogue of the Laplacian).

We note that free I-waves φξ (x) = e(hξx) are eigenfunctions of this op-
erator with eigenvalues λ = |ξ|2p . As in the cases of the Mq , Mξ observables,
there is an infinite family of other eigenfunctions distinct from free I-waves.
These functions are localized on the configuration space (describing fixed
ranges of ideas). The spectrum is discrete: λ = pk , k = 0, ±1, ±2. Thus
the kinetic I-energy is changed only by jumps. The point λ = 0 is the only
point that belongs to the non-discrete spectrum of the operator of the kinetic
I-energy.

87
 Interactions between the brain’s layers as well as interactions with the
external world are described by the operator of the potential I-energy. It is
given by a real-valued function (the potential) V (q1 , ..., qn ). The total mental
energy is represented by the operator:

H = h2 ∆ + V.

We note that the I-potential V (q1 , ..., qn ) can change crucially spectral prop-
erties of the observable of I-energy. If V depends only on p-adic magnitudes
|qj |p of I-coordinates and V → ∞, |qj |p → ∞, and V is bounded from
below (e.g., non-negative), then the spectrum of H (which is discrete) has
only finite degeneracy. Thus the ‘state of mind’ of a free cognitive system
can not be determined by fixing the I-energy. However, by using additional
I-potentials we can (at least in principle) do this.
The ground I-energy state λ0 is not degenerated at all. In the latter case,
by fixing the minimal value of the I-energy H = λ0 we can determine the
‘state of mind’, namely, the λ0 eigenstate. Even for other eigenvalues we can
try to determine the ‘state of mind’ if the degeneracy of the spectrum is not
so large. It is interesting to remark that mathematical results [187] imply
that degeneracy of eigenvalues (distinct from the ground energy) increases
(as p2 ) with increase of p. If we connect the complexity of a cognitive system
with the coding base p then we obtain that for complex cognitive systems
(e.g., p = 1999) it is practically impossible to determine the ‘state of mind’
corresponding to the fixed value of I-energy.
We want to describe the evolution of a quantum-like I-state (I-wave func-
tion) φ(t, x). The first natural and quite nontrivial problem is the choice of
the evolution parameter t, see chapter 6 for an extended discussion of this
problem. There are different natural possibilities for describing the evolution
of I-states: mental time, psychological time, as well as ordinary physical time
evolution. In this chapter we consider the evolution with respect to physical
time t belonging to the real line R, see chapter 6 for models with a mental
evolution parameter.
To derive the evolutional equation for φ(t, x) we proceed in the same way
as Schrödinger in ordinary quantum mechanics. We start with a free I-wave
φξ (x) = e(hξx), ξ, x ∈ Qp . We have:

H0 φξ (x) = |ξ|2p φξ (x),

where H0 = h2 D2 is the operator of the I-energy for a free system. The

φξ (x) is a stationary state corresponding to I-energy E = |ξ|2p . Such a wave
evolves as
φξ (t, x) = eiEt/h φξ (x).

88
We note that this function is a combination of two essentially different expo-
nents: an ordinary exponent and a p-adic character. This function satisfies
the evolutional equation:
∂φ
ih (t, x) = h2 D2 φ(t, x). (52)
∂t
This is Schrödinger’s I-equation for a free cognitive system. If we introduce
an I-potential V (x), As we consider the one-dimensional case, a one-layer
brain, V (x) describes the influence of the external world and the interaction
between neurons in this layer. In the multidimensional case V (x1 , ..., xn )
also describes interaction between distinct layers. then we obtain general
Schrödinger’s I-equation:
∂φ
ih (t, x) = h2 D2 φ(t, x) + V (x)φ(t, x). (53)
∂t
If the initial quantum-like state ψ(x) = φ(0, x) is known then by using (53)
we can find φ(t, x) at each instant t of physical time. Under quite general
conditions [187] the operator H = h2 D2 + V (x) is a self-adjoint operator.
Therefore (53) is the standard Schrödinger’s equation in the Hilbert space H
for one rather special class of operators H. There also are mathematical re-
sults on analytical properties of solutions and correctness of Cauchy problem
[107].
Remark 4.1. (Bohmian theory speculations, cf., chapter 6) In principle
we can define a quantum-like mental potential

h2 2
Vq−l (x) = − D R, where R(t, x) = |φ(t, x)|. (54)
R
However, we can not define the evolution of an individual I-state (position):
q(t). Nevertheless, Vq−l (x) can be considered as a quantative measure of men-
tality.quantative measure of mentality In principle, if we find a quantum-like
state φ(t, x) then we can estimate the ‘level of mentality’ of cognitive sys-
tems under consideration by calculating Vq−l (x). We underline the following
properties of Vq−l (x) :
1) It does not depend on the absolute magnitude of φ;
2) It depends on the second variation of the magnitude of φ.
Thus systems with a slowly varying (in particular, constant) φ(x) have a
low level of mentality; systems with quickly varying φ(x) have a high level
of mentality (in this state φ). We remark that in this approach there are no
jumps in the mentality level. Computers, insects, animals, human beings

89
have levels of mentality corresponding to shapes of φ(x). In principle, if
a computer could approach high magnitudes of Vq−l (x) it would become
conscious ... .
[Bohmian Mechanics on Information Spaces] Bohmian Mechanics on
Information Spaces
In this chapter we develop the formalism of classical information mechan-
ics by analogue to the formalism of the ordinary Newton mechanics which
describes the motion of material systems. We propose information analogues
of Newton’s laws of classical mechanics. Mathematically these laws can be
described by differential equations in I-spaces. Starting with an initial I-
state x0 , we obtain the trajectory q(t) in I-space.
However, classical I-mechanics is not obtained as a copy of the ordinary
classical mechanics. In particular, time t, the parameter of the evolution of
I-states, could not be always identified with physical time tphys . This t is
internal time of an I-system. We can call it I-time (for cognitive systems
– mental or psychological time). The velocity v(t) of the evolution of an
I-state has the meaning of the motivation (to change the I-state q(t) of an
I- system). Formally the velocity is calculated (as in Newton’s mechanics)
as the derivative: v(t) = dq(t)
dt
(on the p-adic space which we use as a model
for I-space). Forces f (t, q) and potentials V (t, q) are information (mental)
forces and potentials which are applied to I-states of I-systems. An I-force
changes the motivation. This change of the motivation implies the change of
the I-state q of an I-system.
Discrete dynamical models were considered in chapters 1-3. In this chap-
ter we study ‘continuous time’ evolution. On one hand, this gives the pos-
sibility to apply (at least formally) the scheme of the standard formalism of
the classical mechanics. On the other hand, in the present model we can
discuss carefully the meanings of I-time and I-velocity (in particular, mental
time and velocity of thoughts).
Classical I-mechanics describes unconscious processes. The phenomenon
of consciousness cannot be explained by the formalism of classical I-mechanics.
To explain this phenomenon, we develop a variant of quantum I-mechanics.
In this model an I-state moves in I-space not only due to classical I-forces,
but also due to a new I-force, namely, quantum I-force. This quantum I-
force (which will be called a conscious force conscious force and denoted by
fC (q)) is induced by an additional I-potential. Such a potential is a quantum
potential on I-space or conscious potential, C(q). The C(q) could not be re-
duced to neurophysiolocal processes in the brain. It is generated by mental
processes. Formally the conscious potential C(q) is calculated with the aid
of the wave function Ψ(q) of an I-system in the same way as in the ordinary

90
pilot wave theory for material systems [34]-[37]:

C = −R00 (q)/R(q), R2 (q) = Ψ(q)Ψ̄(q).

In our model this wave function Ψ is nothing than an I-field, a conscious

fieldconscious field. In the mathematical formalism this field is described
as a function Ψ : XI → XI , where XI is I-space. The evolution of the
Ψ-function is described by an analogue of the Schrödinger equation on the
I-space.
In fact, our formalism of conscious forces and fields is a natural extension
of the well known pilot wave theory (that was developed by De Broglie and
Bohm [56], [34]-[37], [27]) to I-phenomena and, in particular, cognitive phe-
nomena. Even in the pilot wave theory for material systems, especially in
its variant developed in the book of Bohm and Hiley [37], the quantum wave
function Ψ is merely an I-field. However, Bohm-Hiley field of information is
still defined on the real space R3 of localization of material systems. This
I-field acts to material objects. There arises the extremely complex problem
of information-matter interaction which is not yet solved. In our model a
conscious field, Ψ-function, is associated with purely I-processes. It acts to
I-states - collections of hierarchic information strings.
By our model each configuration of I-states of an I-system produces a
new (nonclassical) field, conscious field Ψ. This field induces a new I-force fC
which induces a permanent perturbation of the evolution of I-states in the
I-space. This Ψ-function is nothing than the mathematical representation
of consciousness.consciousness Of course, this formalism is just the first step
to describe the phenomenon of consciousness on the basis of our model of
I-reality. However, even this formalism implies some consequences which
might be interesting for neurophysiology, psychology, artificial intelligence,
complex I-systems, evolutionary biology and social sciences. We present
some of these consequences.
Flows of cognitive information in the brain and other cognitive systems
can be described mathematically in the manner which is similar to the clas-
sical Newton mechanics for motions of material systems. Motion of I-states,
associations, ideas, notions, images in the brain has a deterministic char-
acter. Such a motion is perturbed by various I-noises and controlled by
artificial randomness, see chapters 2,4 for the details. This motion in I-space
is not an evolution with respect to physical time tphys , but with respect to I-
time t. I-potentials can connect different thinking processes in a single brain
as well as in a family of brains. The consciousness cannot be induced by a
physical activity of material structures, for example, groups of neurons. It is
induced by groups of evolving I-states. These dynamical groups of I-states

91
produce a new I-field, conscious field. This field induces a new I-force, con-
scious force, which is the direct analogue of the quantum force in the pilot
wave theory for quantum material systems. This conscious force plays the
great role in I-dynamics in the human brain and other conscious cognitive
systems. As in classical I-mechanics, in quantum I-mechanics conscious po-
tentials can connect thinking processes in different cognitive systems (even
in the absence of physical potentials and forces). Therefore it is possible to
speak about the collective consciousnesscollective consciousness of a group
of cognitive systems, in particular, human individuals. We also remark that
different conscious potentials, conscious Ψ-fields, induce conscious forces fC
of different I-strengths. By our model the magnitude of the consciousness
can be measured. Thus different cognitive systems, in particular, human in-
dividuals, may have different levels of the conscious Ψ-field. By our model
we need not suppose that the consciousness is a feature of only the human
brain. Other cognitive systems, in particular, animals and even nonliving
systems, induce conscious Ψ-fields which via conscious forces fC control or
at least change their cognitive behaviours. From this point of view human
individuals and animals differ only by structures of their conscious Ψ-fields.

27 Newton Laws for Information Processes

First we recall some facts from Newton’s classical mechanics. In Newton’s
model motions of material systems are described by trajectories in a space
Xmat of localization of material systems. In the standard mathematical model
Xmat = R3 or some real manifold. Starting with an initial position q0 , a
material object A evolves along the trajectory q(t) in Xmat . Here t is physical
time. The main task of Newton’s mechanics is to find the trajectory q(t) in
the space Xmat . Let us restrict our considerations to the case in that A has
the mass 1. This can always be done by the choice of the unit of mass. In
this case the momentum p(t) of A is equal to the velocity v(t) of motion. In
the mathematical model the velocity v(t) can be found as v(t) = dq(t)dt
≡ q̇(t).
The velocity need not be a constant. It is useful to introduce an acceleration
a(t) of A which is the velocity of the velocity v(t) : a(t) = dv(t)
dt
≡ v̇(t). The
second Newton law says:
a(t) = f (t, q), (55)
where f (t, q) is a force applied to A. As the mass m = 1 and the momentum,
p = mv, is equal to the velocity, we have

ṗ(t) = f (t, q), p(0) = p0 , t, q, p ∈ Xmat . (56)

92
By integrating this equation we find the momentum p(t) at each instant t of
time. Then by integrating the equation

q̇(t) = p(t), q(0) = q0 , t, q, p ∈ Xmat , (57)

we find the position q(t) of A at each instant t of time.

We develop the formalism of classical I-mechanics by the analogue with
Newton’s mechanics. Instead of the material space Xmat , we consider the I-
space XI , see chapter 1 for the mathematical model based on p-adic trees. An
I-system τ (in particular, a cognitive system) is considered as a transformer
of information: an I-state q ∈ XI of this transformer is in the process of con-
tinuous evolution. A transformer of information τ is continuously performing
transformations
q → q 0 → q 00 → ... .
The time parameter of this evolution is also an I-parameter, namely I-
timeinformation time of τ, t ∈ XI . The activity of τ generates the trajectory
q(t) in the I-space XI . Our deterministic information postulate which is the
generalization of Newton’s deterministic postulate is that the trajectory q(t)
of the evolution of an I-state is determined by forces and initial conditions.
As in Newton’s mechanics, we introduce the velocity v(t) of changing of the
I-state q(t). This is again an I-quantity, a new I-state. It can be calculated
as the derivative in the I-space XI of q(t) with respect to I-time t, see chapter
3 on the p-adic derivative.
We start with development of the formalism for an I-system τ with the I-
mass 1. Here we can identify the velocity v with the momentum p = mv. We
shall call p a motivationmotivation to change the I-state q(t). We postulate
that I-dynamics in XI is described by an I-analogue of Newton’s second law.
The trajectory p(t) of the motivation of τ is described by the equation

ṗ(t) = f (t, q), p(0) = p0 , t, q, p ∈ XI , (58)

where f (t, q) is an I-force which is generated by external sources of informa-

tion; in particular, by other I-systems. If the initial motivation p0 and I-force
f (t, q) are known, then the motivation p(t) can be found at each instant of
I-time t by integration of equation (58). The trajectory q(t) of the evolution
of the I-state can be found by integration of the equation

q̇(t) = p(t), q(0) = q0 , t, q, p ∈ XI . (59)

We recall that in Newton’s mechanics a force f (q), q ∈ Xmat , is said to

be potential if there exists a function V (q) such that f (q) = − dVdq(q) . The
function V (q) is called a potential. We use the same terminology in the

93
I-mechanics. Here both force f and potential V are functions defined on
the space XI . The potential V (q), q ∈ XI , is an I-potential, I-field, which
interacts with an I-system τ. Such fields are classical fields of information,
I-fields.
As we have already mentioned I-time t need not coincide with physical
time tphys . I-time corresponds to the internal scale of an I-process. For
example, for a human individual τ, the parameter t describes ‘psychological
duration’ of mental processes. Our conscious experience demonstrates that
periods of the mental evolution which are quite extended in the tphys -scale
can be extremely short in the t-scale and vice versa. In general instances of
mental time are ideas which denote stages of the I-evolution of an I-system.
We remark that tphys can be also interpreted as a chain of ideas about counts
n = 1, 2, ..., for discrete tphys and about counts s ∈ R, for continuous tphys . In
principle, physical time tphys can be considered as the special representation
for I-time t. However, it is impossible to reduce all I-times to physical time.
Different I-systems, τ1 , ..., τN , and even different mental processes in an I-
system τ have different I-times, t1 , ..., tN . The use of physical time tphys can
be considered as an attempt to construct the unique time-scale for all mental
processes. However, as we have already mentioned this is impossible. In
particular, we could not claim that in general there is an order structure for
t. It can be that instances t1 and t2 of I-time can be incompatible. Thus
a set which is used to represent I-time in a mathematical model cannot be
imagine as a straight line.

28 Bohmian Mechanics for Hierarchical Infor-

mation
First we recall some facts on Bohmian quantum mechanics for material sys-
tems (elementary particles). The formalism of quantum mechanics was devel-
oped for describing motions of physical systems which deviate from motions
described be Newton’s equations (56), (57). For example, let us consider the
well known two slit experiment. There is a point source of light O and two
screens S and S 0 . The screen S has two slits h1 and h2 . Light passes S through
slits and finally reaches the screen S 0 , where we observe the interference rings.
Let us consider light as the flow of particles, photons. Newton’s equations of
motion (56), (57) could not explain the interference phenomenon: classical
forces f involved in this experiment could not rule photons in such a way
that they concentrate in some domains of S 0 and practically cannot appear in
some other domains of S 0 . The natural idea, see Bohm [34]-[37], is to assume

94
that there appears some additional force fQ , quantum force,quantum force
which must be taken into account in Newton’s equations. Thus instead of
equation (56), we have to consider a perturbed equation

ṗ(t) = f (t, q) + fQ (t, q), p(0) = p0 , t, q, p ∈ Xmat . (60)

It is natural to assume that this new force, fQ (t, q), is induced by some field
Ψ(t, q). In the pilot wave theory for material objects it is postulated that this
field Ψ(t, q) can be found as a solution of the Schrödinger equation

h ∂ψ h2 ∂ 2 ψ
(t, q) = (t, q) − V (t, q)ψ(t, q). (61)
i ∂t 2 ∂q 2
Thus each quantum system propagates together with a wave which guides
this particle. Such an approach to quantum mechanics is called a pilot
wavepilot wave theory. Formally there are two different objects: a parti-
cle and a wave. Really there is one physical object: a particle which is
guided by the pilot wave. The Ψ-field associated with a quantum system has
some properties which imply that Ψ(q) could not be interpreted as the ordi-
nary physical field (as, for example, the electromagnetic field). The quantum
force fQ (q) is not connected with Ψ(q) by the ordinary relation f = dΨ(q)dq
.
The ordinary relation between a force f and a potential V implies that scal-
ing V → cV, where c is a constant, implies the same scaling for the force,
namely, f → cf. In the opposite to such a classical relation quantum force fQ
is invariant with respect to the scaling Ψ → cΨ of the Ψ-function. Thus the
magnitude of the Ψ-function (‘quantum potential’) is not directly connected
with the magnitude of quantum force fQ . According to Bohm and Hiley [37],
Ψ(q) is merely an I-field on material space Xmat . For example, in [37] Ψ(q) is
compared with a radio signal which rules a large ship with the aid of an au-
topilot. Here the amplitude of the signal is not important; only information
carried by this signal is taken into account.
Remark 2.1. The pilot wave theory does not give an answer to the
question: Is some amount of physical energy transmitted by the Ψ-field or
not? The book [37] contains an interesting discussion on this problem. It
seems that, despite the general attitude to the information interpretation of
Ψ, they still suppose that Ψ must carry some physical energy. Compared
with the energy of a quantum system, this energy is negligible (as in the
example with the ship). Another interesting consequence of Bohm-Hiley
considerations is that quantum systems might have rather complex internal
structure. Roughly speaking a quantum system must contain some device to
transfer information obtained from the Ψ-field.

95
 From the introduction to this chapter it is clear how we can transform
classical I-mechanics to quantum I-mechanics, conscious mechanics. The
main motivation for such a development of classical I-mechanics is that be-
haviour of conscious systems cannot be described by a classical I-force f.
Behaviour of a conscious I-system strongly differs from behaviour of uncon-
scious I-system (even if both these systems are ruled by the same classical
I-force f ). In general the I-analogue (58) of the second Newton law is vi-
olated for conscious I-systems. As in the case of material systems, it is
natural to suppose that there exists some additional I-force fC (q) acting to
an I-system. We call such a force a conscious force. conscious force This
force perturbs the trajectory of an I-system in the space XI . The (quantum)
conscious trajectory is described by the equation

ṗ(t) = f (t, q) + fC (t, q), p(0) = p0 , t, q, p ∈ XI . (62)

The conscious force fC (t, q) is connected with a Ψ-field, a conscious field, by

the same relation as in the pilot wave formalism for material systems. An
information Schrödinger equation, describes the evolution of the conscious
Ψ-field.

29 Interactions between Information Systems

In the previous two sections we have studied classical and quantum mechani-
cal formalisms for individual I-systems. In this section we consider collective
classical (unconscious) and quantum (conscious) I-phenomena. We start
with classical (unconscious) I-mechanics. Let τ1 , ..., τN be a family of I-
systems with I-spaces XI,1 , ..., XI,N . We introduce I-space XI of this family
of I-systems by setting

XI = XI,1 × · · · × XI,N .

Elements of this space are vectors of I-states q = (q1 , ..., qN ) of individual

I-systems τj . We assume that there exists an I-potential V (q1 , ..., qN ) which
induces information forces fj (q1 , ..., qN ). The potential V is generated by I-
interactions of systems τ1 , ..., τN as well as by external I-fields. The evolution
of the motivation pj (t) and the information state qj (t) of the jth I-system τj
is described by equations:

ṗj (t) = fj (t, q1 , ..., qN ), pj (0) = p0j , (63)

q˙j (t) = pj (t), q(0) = q0j , t, q, p ∈ XI . (64)

96
 This model can be used not only for the description of collective phenom-
ena for a group of different cognitive systems τ1 , ..., τN , but also for a family
of thinking processes in one fixed cognitive system. For example, it is natu-
ral to suppose that the brain contains a large number of dynamical thinking
processors, see chapter 1 for a mathematical model, π1 , ..., πN , which produce
I-states q1 (t), ..., qN (t), related to different mental functions. We can apply
our classical I-mechanics to describe the simultaneous functioning of think-
ing processors π1 , ..., πN . The main consequence of our model is that I-states
q1 (t), ..., qN (t) and motivations p1 (t), ..., pN (t) do not evolve independently.
Their simultaneous evolution is controlled by the I-potential V (q1 , ..., qN ). It
must be underlined that an interaction between thinking modules π1 , ..., πN
has the purely information origin. The potential V (q1 , ..., qN ) need not be
generated by a physical field (for example, the electromagnetic field). A
change of the I-state, qj → qj0 , or motivation, pj → p0j , of one of thinking
processors πj will automatically imply (via the I-interaction V (q1 , ..., qN )) a
change of the I-states and motivations of all other thinking blocks. In princi-
ple no physical energy is involved in this process of the collective I-evolution.
In some sense this is the process of the cognitive (but still unconscious)
self-regulation. Different cognitive systems can have different I-potentials
V (q1 , ..., qN ) which give different types of connections between thinking blocks
πj .
Moreover, the I-potential V can depend on I-time of a cognitive system,
V = V (t, q1 , ..., qN ). Thus at different instances of I-time t a cognitive system
can have different information connections between thinking blocks π1 , ..., πN .
We are going to describe the collective conscious phenomena. Let τ1 , ..., τN
be a family of I-systems. Classical I-motion is described by classical (un-
conscious) I-forcesThroughout this paper we use classical and quantum as
synonyms of unconscious and conscious. In fact, it would be better to use
only the biological terminology. But we prefer to use also the physical ter-
minology to underline the parallel development of mechanical formalisms for
material and mental systems. fj (t, q1 , ..., qN ) by second Newton law for I-
dynamics (58). As in the pilot wave formalism for many particles, there
exists a Ψ-field, Ψ(q1 , ..., qN ), for any family τ1 , ..., τN of I-systems. This field
is defined on the I-space XI = XI,1 × · · · × XI,N , where XI,j is the I-space of
τj . This field generates additional I-forces fj,C (t, q1 , ..., qN ), conscious forces,
and the Newton’s (classical/unconscious) I-dynamics must be changed to
(quantum/conscious) I-dynamics
ṗj (t) = fj (t, q1 , ..., qN ) + fj,C (t, q1 , ..., qN ), j = 1, 2, .., N. (65)
In general the conscious force fj,C = fj,C (t, q1 , ..., qN ) depends on all I-
coordinates q1 , ..., qN , namely I-states of I-systems τ1 , ..., τN . Thus the con-

97
sciousness of each individual I-system τj depends on I-processes in all I-
systems τ1 , ..., τN . The level of this dependence is determined by the form of
the collective Ψ-function. As in the ordinary pilot wave theory [34]-[37], in
our I-model the factorization
N
Y
Ψ(t, q1 , ..., qN ) = Ψj (t, qj )
j=1

of the Ψ-function implies that the conscious force fj,C depends only on the
coordinate qj . Thus the factorization of Ψ eliminates the collective conscious
effect.
As in classical I-mechanics, the above considerations can be applied to a
system of thinking blocks π1 , ..., πN of the individual cognitive system τ ; for
example, the human brain. The conscious field Ψ of τ depends on I-states
q1 , ..., qN of all thinking blocks.

30 Hamiltonian Equations and Active Informa-

tion
As in previous chapters, we use the field of p-adic numbers Qp as the math-
ematical model of I-space: XI = Qp (or Descartes product of such p-adic
trees).
We present an analogue of the Hamiltonian dynamics on I-spaces. As
usual, we introduce the quantity p(t) = q̇(t) = dtd q(t) which is an I-analogue
of the momentum, a motivation. The space Qp × Qp of points z = (q, p)
where q is the I-state and p is the motivation is said to be a phase I-space.
As in the ordinary Hamiltonian formalism, we assume that there exists a
function H(q, p), I-Hamiltonian, on the phase I-space which determines the
motion of τ in the phase I-space:
∂H ∂H
q̇(t) = (q(t), p(t)), q(t0 ) = q0 , ṗ(t) = − (q(t), p(t)), p(t0 ) = p0 . (66)
∂p ∂q
The Hamiltonian H(p, q) has the meaning of I-energy or mental, or psychical
energy, compare with Bergson [29] and Freud [78]. In principle, I-energy is
not directly connected with the usual physical energy.
The simplest Hamiltonian Hf (p) = αp2 , α ∈ Qp describes the motion
of a free I-systemfree I-system τ , i.e., the I-system which uses only self-
motivations for changing of its I-state q(t). Here by solving the system of
the Hamiltonian equations (in the space of analytic functions on the p-adic
space) we obtain: p(t) = p0 , q(t) = q0 + 2αp0 (t − t0 ). The motivation p is the

98
constant of this motion. Thus the free I-system ‘does not like’ to change its
motivation p0 in the process of the motion in the I-space.
In the general case the I-energy is the sum of the I-energy of motivations
Hf = αp2 which is an analogue of the kinetic energy and potential I-energy
V (q):
H(q, p) = αp2 + V (q).
The potential V (q) is determined by fields of information.field of information
In the Hamiltonian framework we can consider interactions between I-
systems τ1 , . . . , τN . These I-systems have I-times t1 , . . . , tN and I-states
q1 (t1 ), . . . , qN (tN ). By our model we can describe interactions between these
I-systems only in the case in that there is a possibility to choose the same
I-time t for all of them. In this case we can consider the evolution of the
system of I-systems τ1 , . . . , τN as a trajectory in the I-space

X I = QN
p = Qp × · · · × Qp , q(t) = (q1 (t), . . . , qN (t)).

The condition of consistency

t1 = t2 = ... = tN = t (67)

plays the crucial role in many psychological experiments. We can not perform
observations for interactions between arbitrary individuals. There must be
a process of learning for the group τ1 , . . . τN which reduces I-times t1 , . . . , tN
to the unique I-time t.
Let us consider a group τ1 , . . . , τN of I-systems with the internal time
t. The dynamics of I-states and motivations is determined by the I-energy;
H(q, p), q ∈ QN N
p , p ∈ Qp . It is natural to assume that

N
X
H(q, p) = αj p2j + V (q1 , . . . , qN ), αj ∈ Qp .
j=1

PN 2
Here Hf (p) = j=1 αj pj is the total energy of motivations for the group
τ1 , . . . , τN and V (q) is the potential energy. As usual, to find a trajectory
in the phase space QN N
p × Qp , we have to solve the system of Hamiltonian
equations:
∂H ∂H
q̇j = , ṗj = − , qj (t0 ) = q0 , pj (t0 ) = p0 .
∂pj ∂qj

Remark 4.1. (Active information) Our ideas about information and

information field are similar to the ideas of Bohm and Hiley [37], see especially
p. 35-38. As Bohm and Hiley [37], we do not follow

99
 “... Shannon’s ideas that there is a quantative measure of information
that represents the way in which the state of a system is uncertain for us.”
We also consider information as an active information.active information
Such information interacts with I-systems. As a consequence of such inter-
actions I-systems produce new information. The only distinguishing feature
is that material objects are not involved in our formalism. According to
[37] active information interacts with material objects (for example, the ship
guided by radio waves). Bohm and Hiley assume that I-fields have nonzero
physical energy which directs other (probably very large) physical energy.
However, physical energies are not involved in our model. Thus we need not
assume that I-fields have some physical energy. In particular, we need not
try to find, as in [37], p.38, an origin of such an energy. We also remark that
Bohm and Hiley discuss only quantum Ψ-fields. We shall use both classical
and quantum I-fields.
Bohm and Hiley discussed a difference between ‘active’ and ‘passive’ in-
formation. In fact, our model supports their conclusion that ‘all information
is at least potentially active and that complete passivity is never more than
an abstraction...’, [37], p.37. If an I-system τ moves in the field of forces Ψ
(classical or quantum), then the information x ∈ supp V is active for τ and
the information x ∈ Qm p \ supp V is passive for τ. Let v = V (t, x) be a time
dependent potential. Then the set of active information X(t) = supp V(t)
evolves in I-space. Thus some passive information becomes active and vice
versa.

31 Information Mass
It must be underlined that in p-adic analysis the condition f ≡ 0 does not
imply that a differentiable function f is a constant, see chapter 3. There exist
complicated continuous motions z(t) = (q(t), p(t)) in the phase space for
systems having zero I-energy (q̇ ≡ 0 or ṗ ≡ 0). However, such mathematical
pathologies can be eliminated by the consideration of analytical functions. If
f 0 ≡ 0 and f is analytic then f = constant.
We have considered dynamics of I-systems of the unit mass. There the
coefficient v of proportionality between the variation δq of the I-state and
the variation δt of I-time t: δq = vδt, was considered as a motivation. In the
general case the motivation p may not coincide with v. Let us assume that
the motivation p is proportional to v, p = mv, m ∈ Qp . This coefficient m of
proportionality is called an information mass, I-mass.mass We also call v an
I-velocity. Thus δq = mp δt.

100
 Let τ1 and τ2 be two I-systems with the I-masses m1 and m2 and let
|m1 |p > |m2 |p . Let τ1 and τ2 have the variations δt1 , δt2 of I-time of the same
p-adic magnitude, |δt1 |p = |δt2 |p . Let these variations generate the variations
δq1 and δq2 of their I-states of the same p-adic magnitude, |δq1 |p = |δq2 |p .
To make such a change of the I-state, τ1 needs a larger motivation:
δq δq
|p1 |p = | |p |m1 |p > |p2 |p = | |p |m2 |p .
δt δt
Thus the I-mass is a measure of an inertia of information. We define kinetic
1 2
I-energy by T = 2m p . The variation δt of I-time t implies also a variation
δp of the motivation p:
δp = f δt .
The coefficient f of proportionality is called an information force, I-force.force
Thus any change of the motivation is due to the action of an I-force f . If
f = 0 then δp = 0 for any variation δt of t. Thus an I-system cannot change
its motivation in the absence of I-forces (we consider analytic trajectories in
the p-adic space). By analogue with the usual physics we call the coefficient
a of a proportion between the variation δv of the I-velocity v and the vari-
ation δt of the I-time t, δv = aδt, an I-acceleration. Thus δp = amδt. This
relation can be rewritten in the form of an I-analogue of the second Newton
law:
ma = f or ṗ = f. (68)
The total I-energy H is defined as the sum of the kinetic and the potential
1 2
I-energies, H(q, p) = 2m p + V (q). The Hamiltonian equation ṗ = − ∂H ∂q
coincides with the Newton equation ṗ = f .
As it was already noticed, there exist complicated motions with zero I-
energy. In psychological models these motions can be interpreted as motions
without any motivation. Such motions do not need I-force. On the other
hand, we can consider an I-potential V (q) such that ∂V
∂q
= 0. The potential
I-energy V (q) can have complicated behaviour on the I-space X = Qp . At
the same time the I-force f = 0. Thus there may exist I-fields which do not
induce any I-force.
In psychological models we can interpret analytical trajectories in the
phase I-space as a normal behaviour, i.e., an individual needs a motivation
for the change of a psychological state. We can observe some psychologi-
cal (information) force which induces this change. There is a psychological
(information) field which generates this force. Trajectories (non-analytical)
with zero motivation are interpreted as abnormalabnormal psychological be-
haviour. Probably such trajectories correspond to mental diseases. On the

101
other hand, they may explain anomalous phenomena.anomalous An individ-
ual changes his psychological state without any motivation in the absence of
any psychological (information) force. Here, in fact, a p-adic generalization
of the Hamiltonian formalism does not work. We need to propose a new
physical formalism to describe such phenomena.

32 Wave Functions Taking Values in p-adic Fields

It is natural to quantize classical mechanics on I-spaces over Qp . We give the
following reasons for such quantization. Observations over I-quantities are
statistical observations. We have to study statistical ensembles of I-systems,
instead of studying of an individual I-system. Such statistical ensembles are
described by quantum states φ. As usual in quantum formalism, we can
assume that a value λ of an information observable, I-observable, B can be
measured in the state φ with some probability Pφ (B = λ). This ideology
is nothing than the application of the statistical (ensemble) interpretation of
quantum mechanics (see, for example, Ballentine [21], [22]) to the information
theory. By this interpretation any measurement process has two steps: (1) a
preparation procedure E; (2) a measurement of a quantity B in the states φ
which were prepared with the aid of E.
Let us consider these steps in the information framework. A preparation
procedure E we have to select a statistical ensemble φ of I-systems on the
basis of some I-characteristics. Typically in physics for material objects a
preparation procedure E is realized as a filter based on some physical quantity
A, i.e., we select elements which satisfy the condition A = µ where µ is one of
the values of A. We can do the same in quantum I-theory. An I-observable A
is chosen as a filter, i.e., I-systems for the statistical ensemble φ are selected
by the condition A = µ where µ ∈ Qp is some information. For example, we
can choose A = p, the motivation, and select a statistical ensemble φ = φ(p =
µ) of I-systems which have the same motivation µ ∈ Qp . Then we perform
the second step of a measurement process and measure some I-observable B
in the state φ(p=µ) . For example, we can measure the I-state q of I-systems
belonging to the statistical ensemble described by φ(p=µ) . We shall obtain a
probability distribution P(q = λ|p = µ), λ, µ ∈ Qp (a probability that I-
system has the I-state q = λ under the condition that it has the motivation
p = µ). It is also possible to measure the I-energy E of I-systems. We shall
obtain a probability distribution P(E = λ|p = µ), λ, µ ∈ Qp . On the other
hand, we can prepare a statistical ensemble φ(q=µ) by fixing some information
µ ∈ Qp and selecting all I-systems which have the I-state q = µ. Then we
can measure motivations of these I-systems and we shall obtain a probability

102
distribution P(p = λ|q = µ).
Another possibility is to use a generalization of the individual interpre-
tation of quantum mechanics, so called orthodox Copenhagen interpretation
[191], [64], [60], [151]. By this interpretation a wave function Ψ(x), x ∈ Rn ,
describes the state of an individual quantum particle. In the same way we
may assume that a wave function Ψ(x), x ∈ Qnp , on the I-space describes
the state of an individual I-system τ The problem of interpretations is the
important problem of ordinary quantum mechanics on real space. The same
problem arises immediately in our quantum I-theory. We do not like to start
our investigation with a hard discussion on the right interpretation. We can
be quite pragmatic and use both interpretations by our convenience..
In chapter 5 we presented the mathematical model of quantum I-mechanics
based on wave functions φ : Qp → C, where C is the ordinary field of com-
plex numbers. In this section we shall present another mathematical model
of quantum I-mechanics based on wave functions φ : Qp → Qp . Both mod-
els, namely I-mechanics with C-valued and Qp -valued wave functions, have
advantages and disadvantages (mathematical, physical and cognitive). As
we have already seen, the main disadvantage of the model with C-valued
wave functions is impossibility to introduce I-observables corresponding to
the I-state q of an I-system τ and the motivation p to change this I-state.
In the model with Qp -valued wave functions we can find a representation
of the I-state and motivation observables by symmetric operators q̂ and p̂
in the p-adic Hilbert space. However, the later formalism induces nontrivial
problem with Born’s probability interpretation of a wave function, see further
considerations. At the moment we could not say what formalism provides
the right mathematical description of measurements of I-observables.
In fact, a mathematical model for quantum I-formalism has been already
constructed. This is quantum mechanics with p-adic valued functions, see
[105], [107], [109]-[111], [2]. We present briefly this model. The space of
quantum states is realized as a p-adic Hilbert space K [107], [2]. This is a Qp -
linear space on which is defined a symmetric bilinear form (·, ·) : K×K → Qp .
This form is called an inner product on K. By definition quantum I-state
φ is an element of K such that (φ, φ) = 1; quantum I-observable A is a
symmetric bounded operator A : K → K, i.e., (Ax, y) = (x, Ay), x, y ∈ K.In
p-adic models we do not need to consider unbounded operators, because all
quantum quantities can be realized by bounded operators, see [111]. We
discuss a statistical interpretation of quantum states in the case of a discrete
spectrum of A. Let {λ1 , . . . , λn , . . .}, λj ∈ Qp be eigenvalues of A, Aφn =
λn φn , φn ∈ K, (φn , φn ) = 1. The eigenstates φn of A are considered as pure
quantum I-states for A, i.e., if an ensemble of I-systems is described by the
state φn then the I-observable A has the value λn ∈ Qp with probability 1.

103
Let us consider a mixed state
∞
X
φ= qn φn , qn ∈ Qp , (69)
n=1

where (φ, φ) = ∞ 2
P
n=1 qn = 1.As in the usual theory of Hilbert spaces, eigen-
vectors corresponding to different eigenvalues of a symmetric operator are
orthogonal. If we perform a measurement of the I-observable A for systems
belonging to the statistical ensemble described by φ then we obtain the value
λn with probability P (A = λn |φ) = qn2 .
As we have already discussed, the main problem (or the advantage?) of
this quantum model is that these probabilities belong to the field of p-adic
numbers Qp . The simplest way is to eliminate this problem by considering
only finite mixtures (69) for which qn ∈ Q (we remark that the field of rational
numbers Q is a subfield of Qp ). In this case the quantities P(A = λn |φ) = qn2
can be interpreted as usual Kolmogorov probabilities [126]. Therefore we may
assume that there exist (can be prepared) quantum I-states φ which have the
standard statistical interpretation: when the number N of experiments tends
to infinity, the frequency νN (A = λn |φ) of an observation of the information
λn ∈ Qp tends to the probability qn2 .
However, we can use a more general viewpoint to this problem. Non-
Kolmogorov probability model with p-adic probabilities has been developed
in [106], [107], [108], [111], [112]. If we use a p-adic generalization of a
frequency approach to probability, see von Mises [143], then p-adic probabil-
ities are defined as limits of relative frequencies νN with respect to the p-adic
topology. It is quite surprising that in the p-adic framework we can obtain
negative rational frequency probabilities [106], [107], [111], [117]. On the
other hand, negative ‘probabilities’ appear in many quantum models (see,
for example, [112]). P. A. M. Dirac was the first who introduced explicitly
the concept of negative probability (in close connection with the concept
of negative energy), [63]. R. Feynman also discussed the possibility to use
negative probabilities in quantum formalism, see [74]. In particular, he re-
marked: ‘The only difference between a probabilistic classical world and the
equations of the quantum world is that somehow or other it appears as if the
probabilities would have to go negative, and that we do not know, as far as
I know, how to simulate.’
By using the p-adic frequency probability model for the statistical in-
terpretation of quantum I-states we may assume that there exist I-states φ,
ensembles of I-systems, such that the relative frequencies νN (A = λn |φ) have
no limit in R. We cannot apply the standard law of the large numbers in
this situation. If we perform measurements of an I-observable A for such a

104
quantum state and study the observed data by using the standard statistical
methods, then we shall not obtain a definite result. There will be only ran-
dom fluctuations of relative frequencies, see [107]. Such a behaviour can be
demonstrated in psychological experiments. The possibility of the use of p-
adic probability models gives the important consequence for scientists doing
experiments with a statistical I-data: the absence of the statistical stabiliza-
tion, random fluctuation, does not imply the absence of an I-phenomenon.
This statistical behaviour may have the meaning that this I-phenomenon
cannot be described by the standard Kolmogorov probability model.
The evolution of a p-adic wave function is described by an I-analogue of
the Schrödinger equation:
hp ∂Ψ h2p ∂ 2 Ψ
(t, x) = (t, x) − V (t, x)Ψ(t, x), (70)
i ∂t 2m ∂x2
where m is the I-mass of a quantum I-system. Here a constant hp plays
the role of the Planck constant. By purely mathematical reasons (related to
convergence of p-adic exponential and trigonometric series, see chapter 3, see
also [107], [111]) it is convenient to choose hp = p1 .
√
The use of the imaginary unit i = −1 in equation (70) implies the con-
sideration of the quadratic extension Qp (i) = Qp × iQp of Qp . Elements of
this extension have the form z = a + ib, a, b ∈ Qp . This extension is well
defined for p = 3, mod 4, see [107], [111]. As usual, we introduce the con-
volution z = a − ib. Here we have zz = a2 + b2 . In what follows we assume
that wave functions take values in Qp (i) = Qp × iQp . It would be noticed
that we use this quadratic extension of Qp only by one reason: to get expres-
sions, in particular, formulas for solutions of differential equations, which are
formally coincide with expressions derived the ordinary quantum formalism.
The formalism of p-adic quantum mechanics has been developed for arbitrary
quadratic extensions of Qp , as well as other non-Archimedean fields, [111];
see also [107] on a quantum formalism over nonquadratic extensions.
Example 6.1. (A free conscious system). Let the potential V = 0.
Then the analytic solution of the Schrödinger equation corresponding to the
p2
I-energy E = 2m has the formWe note that formal expressions for analyti-
cal solutions of p-adic differential equations coincide with the corresponding
expressions in the real case (in fact, we can consider these equations over
arbitrary number field, see [107]). However, behaviours of these solutions are
different. :
Ψp (t, x) = ei(px−Et)/hp . (71)
By the choice hp = 1/p this function is well defined for all x ∈ Qp and
t ∈ Qp . As R2 (x) = ΨΨ ≡ 1, this wave function describes the uniform (p-

105
adic probability) distribution, see [107], [111] on the ring of p-adic integers
Qp . This is a Qp -valued ‘Haar measure’ on Qp . Thus a I-system τ in the
state Ψ can be observed with equal probability in any state x ∈ Qp . In this
sense behaviour of a free I-system is similar to behaviour of the ordinary free
quantum particle. On the other hand, there is no analogue of oscillations.
Of course, as usual,

Ψp (t, x) = cos(px − Et)/hp + i sin(px − Et)/hp .

However,

| cos(px − Et)/hp |p = 1, | sin(px − Et)/hp |p = |(px − Et)/hp |p .

33 Information Waves on p-adic Trees

We now consider the pilot wave version of the p-adic quantum I-mechanics.
Let us consider a system of N I-systems, τ1 , ..., τN , with Hamiltonian
X p2 X
k
H= + Vki (xk − xi ).
k
2mj k>i

The wave function Ψ(t, x), x = (x1 , ..., xN ), xk ∈ Qmp (where m is the dimen-
sion of I-space which is used for the description of the I-state of τj ) evolves
according to Schrödinger equation (70). A purely mathematical consequence
of this is that
∂ X ∂
ρ(t) + jk (x, t) = 0, (72)
∂t k
∂xk

where ρ(t, x) = Ψ(t, x)Ψ(t, x) is a probability density on the configuration

I-space QmN
p and
∂
jk (x, t) = m−1
k Im(Ψ(t, x) Ψ(t, x)).
∂xk
We assume that a quantum I-system τk has at any instant t of I-timeOf
course, we assume that I-times t1 , ..., tN of I-systems τ1 , ..., τN satisfy the
condition of consistency (67). well defined I-state xk and motivation pk . The
evolution of the I-state xk is described by the differential equation:
jk (t, x)
ẋk (t) = . (73)
ρ(t, x)
The derivation of this equation is performed in the same way as in the stan-
dard pilot wave theory for material systems [37]. From the mathematical

106
point of view the only difference is that we have to use the differential calcu-
lus on p-adic spaces, instead of real spaces. As we have already remarked, to
perform standard calculations on p-adic spaces, we have to consider analytic
functions.
This model gives the natural description of an evolution of the I-states of
a system of I-systems τ1 , ..., τN . We consider a system S of brains τ1 , ..., τN .
The wave function Ψ(t, x) of the S depends on I-states of all brains in the S.
Thus motions of these brains in the phase I-space Q2m p are not independent.
QN
If, as in ordinary real formalism, Ψ(t, x) = j=1 Ψj (t, x), then the I-motions
of brains τj are independent. In such a case there are no correlations between
consciousness of different brains.
Remark 7.1. (Non-locality) We discuss the problem of non-locality of
the pilot wave formalism. Often non-locality is considered as one of the
main difficulties of the pilot wave formalism. However, non-locality is not a
difficulty in our pilot I-wave formalism. This is non-locality in I-spaces. Such
non-locality can be natural for some I-systems. For I-systems, information
non-locality means that ideas which are separated in a p-adic space can be
correlated. However, p-adic separation means only that there are no strong
associations between ideas or groups of ideas. But this absence of associations
does not imply that these ideas could not interact.
By our model each human society S has a wave function Ψ(t; x). The same
considerations can be applied to animals and plants. The only difference is
probably that in the latter cases quantum I-potentials are not so strong.
Thus we get the conclusion that there may exist a wave function Ψliv (t, x) of
all living organisms. The wave function Ψ(t, x)liv (t, x) can be represented in
the form: X
Ψ(t, x)liv (t, x) = Ψf (t, x), (74)
f

where Ψf (t, x) is a wave function of the living form f. An observable F, that

values are living forms f, can be realized as a symmetric operator in a p-adic
Hilbert space, F Ψf = f Ψf , where f ∈ Qp is the code of the living form f
in the alphabet {0, 1, ..., p − 1}. By equation (73) the evolution of the fixed
form f0 depends on evolutions of all living forms f.
The process of the evolution of living forms is not just a process based on
Darwin’s natural selection. This is a process of a quantum I-evolution in that
the conscious field of all living forms plays the important role. This model
might be used to explain some phenomena which could not be explained
by Darwin’s theory.Darwin For example, the beauty of colours of animals,
insects and fishes could not be a consequence only of the process of the
natural evolution. This is a consequence of the structure of the conscious

107
field Ψ(t, x)liv (t, x) Our model implies the existence of such a field. Therefore,
the process of colours’ evolution is a process of the simultaneous evolution of
colours of numerous living forms. These colours do not serve only to improve
fitness of concrete forms, as it should be due to Darwin’s theory, but they were
produced by the correlated evolution of all living forms. Such an evolution
of colours improves ‘fitness’ of the whole system.. By the same reasons we
can explain some aspects of relations between robbers and victims. It seems
that nature is the well organized system which gives robers a possibility to
eat victims. This system is nothing else than a result of the evolution due to
(73).
The field Ψ(t, x) which gives the mathematical representation of the con-
sciousness of a family of cognitive systems τ1 , ..., τN can be considered as a
new cognitive system, T. T gets information from cognitive systems τ1 , ..., τN
via the classical field V. T changes I-states of τ1 , ..., τN via equation (73).
The only distinguishing feature is that the I-state of T cannot be identi-
fied with a point in the I-space Qm p for a finite m. However, if we extend
the I-formalism by using infinitely dimensional I-spaces over Qp , see [107],
then T can be considered as a cognitive system. Thus each cognitive sys-
tem or a group of cognitive systems induce a new cognitive system T, the
consciousness, which evolves in an infinitely dimensional I-space.
In principle, T may induce a new field Φ(t, Ψ(·)), namely a conscious field
of the conscious field Ψ. This field determine a quantum potential for T. The
field Φ(t, Ψ(·)) can be again considered as a cognitive system T 1 which evolves
in infinite dimensional I-space. It is the consciousness of the consciousness
T . Such a construction can be repeated infinitely many times. There is a
‘conscious tower’, T, T 1 , ..., T n , ... . We might speculate that the motion of
a cognitive system in I-space is determined by the hierarchic conscious sys-
tem T, T 1 , ..., T n , ... . Of course, the effect of T 1 is not present in the linear
Schrödinger equation. If we accept the hypothesis on the hierarchic con-
scious structure, then the linear Schrödinger equation has to be changed to
nonlinear equation. Hence the cognitive considerations support de Broglie’s
ideas about nonlinear perturbations of Schrödinger’s equation [56]. By de
Broglie’s theory for material systems a quantum particle is described as just
a singularity in the guide wave. We do not plan to develop de Broglie’s the-
ory for cognitive systems in this book. The basic postulate of such a theory
would be
Mind is a singularity in the pilot wave on I-space.

108
34 p-adic Bohmian Mechanics and Waves of
Brain Activation
The reader may ask: Why do we need to use an I-analogue of quantum
mechanics to describe conscious phenomena? Why is it not sufficient to use
only classical I-dynamics with Newton’s I-equation, (58)? There are two
reasons to introduce a conscious field ψ(t, q) :
(D) Our conscious experience demonstrates that I-motion of a conscious
cognitive system τcons could not be described by classical Newton’s I-equation
(58). Classical I-forces do not provide the right balance of guiding forces for
the I-state q(t) of τcons . One of possibilities to perturb classical I-equation,
(58), of I-motion is to use an analogue with material objects. These objects
can be considered as a particular class of transformers of information. By
the analogue we introduce a new (conscious) force via the ψ-function which
satisfies Schrödinger I-equation, (70). Of course, such an approach would
be strongly improved if we could explain the mechanism of generation of the
conscious ψ-field by τcons . We note that such a mechanism is still unknown
in the ordinary pilot wave theory for material objects [37].
(S) It is impossible to perform a measurement of the I-state q as well
as the motivation p or some other I-observable for a conscious system τcons
without to disturb this systemThe process of an I-measurement is an interac-
tion of τcons with some I-potential V (t, q). Any τcons is extremely I-unstable.
Even low I-potentials V (t, q) disturb τcons . Such an I-instability is the com-
mon feature of conscious and quantum systems.. In particular, we could not
perform a measurement of both q and p for the same τcons . As in the or-
dinary quantum mechanics, we have to use large statistical ensembles S of
conscious systems and perform statistical measurements for such ensembles
to find probabilities for measurements of I-quantities. One of possibilities is
to use again an analogue with the ordinary quantum formalism and intro-
duce a field of probabilities ψ(t, q) which describes statistical properties of
the ensemble S such that the square of the amplitude of ψ

R(t, q)2 = ψ(t, q)ψ̄(t, q)

gives the probability to find τcons ∈ S in the I-state q at the instant t of

I-time.
One of the main features of the pilot wave theory for material objects
is that the pilot wave field coincides with the probability field: ψ(t, x) =
ψpilot (t, x) = ψ(t, x)prob . We also postulate that the conscious ψ-field which
generates the conscious force fC coincides with the probability field.
Let τcons be a conscious system. The performance of the I-state q = q(t)

109
of τcons can be viewed as the performance of an image on some screen S. This
screen continuously demonstrates ideas of τcons . The τcons is a self-observer
for these ideas. As we have already noted, our conscious experience says that
a new I-state q(t + ∆t), an image on S at the instant t + ∆t, is generated not
only on the basis of the previous I-state q(t), an image on S at the instant t.
The main feature of τcons is that q(t+∆t) depends on all information which is
collected in τcons .Hiley and Pylkkännen [92] call such a property ‘the unity of
consciousness’. Thus Newton’s I-equation (58) must be modified to describe
such a unity of information in the process of the I- evolution of τcons .
Denote by D(t) ≡ Dτcons (t) the domain in XI corresponding to informa-
tion which is contained in the memory of τcons at the moment t. Then

q(t + ∆t) = F (q(t), D(t)),

where F is a transformation that depends on the whole information in D(t).

The main problem is to find a transformation F which can provide the ad-
equate description of the conscious evolution of an I-state. We propose the
following model.
For each I-state x ∈ D(t), we define an I-quantity ψ(t, x) which de-
scribes the I-activity of the I-state x. In this way we introduce a new I-field
ψ(t, x), x ∈ D(t), a field of memory activation.memory activation It is postu-
lated that the field of memory activation is represented in our mathematical
model by the pilot wave ψ(t, x), conscious field, which guides the I-state q(t)
of τcons .
Let t and t0 be two different instances of time. Schrödinger I-equation
(70) implies that ψ(t0 , x), can be found as the result of integration of ψ(t, x)
over the memory domain D(t) :
Z
0
ψ(t , x) = K(t0 , t, x, y)ψ(t, y)dy, (75)
D(t)

where the kernel K(t0 , t, x, y) describes the time propagation of memory ac-
tivation. K(t0 , t, x, y) describes the influence of an I-state y ∈ D(t) on the
I-state x ∈ D(t0 ). Here dy is a Qp -valued Haar distribution on Qp , [107].
One of the main consequences of the quantum I-formalism is that mem-
ory activation fields at different instances of time are connected by a linear
transformation.
We recall that the Bohmian mechanics implies that the quantum force fQ
does not depend on the amplitude R of a quantum field ψ. Roughly speaking
the strength of fQ depends on variations of R :
fQ = −Q0 ,

110
 where a quantum potential Q is defined as

Q = −R00 /R .

For example, if R ≡ Const, then fQ = 0; if R(x) = x (rather slow variation),

then fQ is still equal to zero; if R(x) = x2 , then fQ 6= 0.
As we have remarked a few times, the differential calculus in spaces of
analytic functions on XI = Qp reproduces calculations that as we have in
the real case; in particular, the form dependence, for conscious potentials
C(x) = −R00 /R (in the I-framework we use the symbol C(x) instead of Q(x)
to underline conscious interpretation of this quantity).
Moreover, we provide the explanation of this variation dependence of
C(x) on R(x). In our model the pilot information ψ-function is interpreted
as a memory activation field. Our conscious experience says that the uni-
formly high activation of all I-states in the memory of τcons could not imply
a conscious behaviour. The uniform activation, R(x) = Const, x ∈ Dτcons ,
eliminates at all a memory effect from the evolution of the I-state q(t) of τcons .
Only a rather strong variation of the amplitude R(x), x ∈ Dτcons , of field of
memory activation produces a conscious perturbation of the I-motion of τcons .
Of course, the information pilot wave theory predicts essentially more
than we can extract from our vague conscious experience. As in the Bohmian
mechanics,
fC = (R000 R − R00 R0 )/R2 .
Thus the conscious force fC depends not only on the first variation δR of the
amplitude of the memory activation field, but also on the second and third
variations, δ 2 R, δ 3 R. These predictions might be verified experimentally.
Remark 8.1 (Neurophysiologic links). In the spatial neurophysiologic
models of memory spatially extended groups of neurons represent human
images and ideas. Thus it is possible to construct a transformation Js :
Bs → Bi , where Bs ⊂ Xmat = R3 is the spatial domain of a ‘physical brain’
and Bi ⊂ XI = Qkp is the I-domain of an ‘information brain.’ Thus the
memory activation field ψ(x), x ∈ Bi , can be represented as an I-field on the
spatial domain Bs : φ(u) = ψ(Js (u)). We now consider the simplest model
with the 0/1-coding (nonactivated/activated) of activation of I-states x in
the memory. Roughly speaking the information pilot wave theory predicts
that only large spatial variations of φ(u) can imply conscious behaviour. Of
course, this is a vague application of our (information) pilot wave formal-
ism. The transformation Js can disturb the smooth structure of the I-space
Qkp and variations with respect to neuron’s location u ∈ Bs ⊂ R3 may
have no links to variations with respect to I-state’s location x ∈ Bi ⊂ Qkp .

111
Moreover, the purely mathematical experience says that this is probably the
case. The most natural transformations a : Qp → R have images of fractal
types, [187] . This can be considered as a reason in favour of the frequency
domainfrequency domain model,[94], by that dust-like configurations of neu-
rons in Bs oscillating with the same frequency seem to correspond to the
same idea, image. In the latter model it is possible to construct a transfor-
mation Jf : Bf → Bi , where Bf ⊂ R3 is the domain of a ‘frequency brain’.
Thus the field ψ(x), x ∈ Bi , can be represented as an I-field on the frequency
domain Bf : φ(ν) = ψ(Js (ν)). The above experimental predictions can be
also tested for variations of φ(ν).
Example 8.1. (A free conscious system). Let us consider the memory
activation field of a free conscious system τcons . Suppose that the memory
of τcons is activated by some concrete motivation, p = α which is not mixed
with other motivations. Then ψα (x) = eiαx/hp , where hp = 1/p and p > 1
is a prime number (the basis of the coding system of τcons ). Such a field
can be called a motivation wave. This motivation wave propagates via the
Schrödinger I-equation:
ψα (t, x) = exp{i(αx − Eα t)/hp },
where Eα = α2 /2m is the information (psychical) energy of the motivation
p = α. Here S(t, x) = (αx − Eα t) and R(t, x) ≡ 1. Thus the conscious force
fC ≡ 0.
Suppose now that two different motivations, p = α and p = β, activate
the memory of τcons . The corresponding motivation waves are ψα (x) = eiαx/hp
and ψβ (x) = eiβx/hp . Suppose that these waves have amplitudes dα , dβ ∈ Qp .
Suppose also that there exists a phase shift, θ, between these two waves of
motivations in the memory of τcons . One of consequences of the quantum
I-formalism is that the total memory activation field ψ(t, x) is a linear com-
bination of these motivation waves:
ψ(x) = dα eiθ/hp ψα (x) + dβ ψβ (x).
The presence of the phase θ implies that the motivation p = α started
to activate the memory earlier than the motivation p = β, namely at the
instant s = −θ/Eα . If the motivation p = α has a small I-energy, namely,
|Eα |p << 1, then a nontrivial phase shift θ can be obtained for a rather large
time shift s. The I-motion in the presence of two competitive motivation
waves in the memory of τcons is quite complicated. It is guided by a non-
trivial conscious force fC (t, x). We omit rather complicated mathematical
expression which formally coincides with the standard expression. As the
memory activation field coincides with the field of probabilities, probabili-
ties to observe motivations p = α and p = β are equal to d2α and d2β . As

112
it was already mentioned, in general these are not rational numbers. Thus
in general these probabilities could not be interpreted as ordinary limits of
relative frequencies. There might be violations of the law of large numbers
(the stabilization of frequencies) in measurements on conscious systems.
Complexity of the I-motion essentially increases if ψ is determined by
k ≥ 3 different motivations. Finally we remark that (in the opposite to
the Bohmian mechanics) waves ψα and ψβ , α 6= β, are not orthogonal. The
covariation < ψα , ψβ >6= 0. Thus all motivation waves in a conscious system
are correlated.
Example 8.2. (Conscious evolution of complex biosystems) Let τ be
a biosystem having a high I-complexity and let l1 , ..., lM be different living
forms belonging to τ. We consider the biosystem τ as an I-object, trans-
former of information, with the I-state q τ (t). It is supposed that τ has a
kind of collective memory. Let ψ τ (t, x) be the memory activation field of
τ. I-dynamics of τ depends not only on classical I-fields, but also on the
activation of the collective memory of τ. Suppose that τ can be considered
(at least approximately) as an I-isolated biosystem. Each living form lj has
a motivation αj to change the total I-state q τ . The pilot I-formalism implies
that the total motivation pτ of the biosystem τ could not be obtained via the
summation of motivations αj . The mechanism of generation of pτ is more
complicated. Each αj activates in the collective memory of τ the motivation
wave ψαj . A superposition of these waves gives the memory activation field,
conscious field, of τ
X
ψ τ (t, x) = dj eiθj /hp ψαj (t, x),
j

compare with (74). For some biosystems, amplitudes dj , j = 1, ..., M, can be

chosen as sizes of populations of lj , j = 1, ..., M. The ψ τ (t, x) induces rather
complicated conscious potential C τ (t, x) which guides the motivation pτ and
I-state q τ of τ.
Remark 8.2. (Materialistic axiom) Finally, we discuss the correspon-
dence between states of brain and states of mind. The thesis that to every
state of brain there corresponds a unique state of mind is often called the
materialistic axiom of cognitive science,materialistic axiom see Bergson [29].
Let us consider again a free conscious system τcons . It will be shown that
motivations of τcons could not be identified with waves of memory activation.
Suppose that there exists a fixed motivation p = α which activates the
memory of τcons . We know that the field ψα (x) = eixα/hp is the eigenfunction
of the position operator p̂. Thus, for an ensemble Sα of free systems with
the same memory activation field ψα (x), observations of the motivation will

113
give the value p = α with the probability 1. Let λ(x) and θ(x) be arbitrary
differentiable functions, Qp → Qp , having zero derivatives and such that all
exponential functions under consideration are well defined, see chapter 3.
Set ψαλ,θ (x) = R(x)eiS(x)/hp , where R(x) = eλ(x) and S(x) = αx − θ(x). The
ψαλ,θ (x) is also an eigenfunction of the motivation operator. Thus observations
of the motivation for an ensemble Sαλ,θ of conscious systems with memory
activation field ψαλ,θ (x) will also give the motivation p = α. However, the fields
ψαλ,θ (x) and ψα (x) can have extremely different distributions of activation of
I-states in the memory. By any external observer all these fields, states
of brain, are interpreted as the same state of mind, namely the motivation
p = α.

35 Conservation Laws
In this section we turn back to classical mechanics on I-spaces. Let I-space
XI is represented as Qp . As usual, we consider only I-quantities described by
analytical functions. As it has been already remarked, we do not claim that
only analytical functions describe real information processes. We like only to
simplify mathematical considerations. It must be also noticed that, in fact,
in the I-framework the difference in behaviours of classical and quantum
systems is not so crucial as in the standard material framework. We start
with consideration of a simple example of a classical I-system, Hooke’s I-
system, which demonstrates some features of quantum I-systems.
Let the I-force f be proportional to the I-state q, f = mβ 2 q, where m is
the I-mass and β ∈ Qp is a coefficient of the interaction. Here (55) gives the
2 p2 2 2
equation q̈ = β 2 q. As f = − ∂V
∂q
, V (q) = − mβ2 q 2 and H(q, p) = 2m − mβ2 q ;
the Hamiltonian equations are q̇ = p/m and ṗ = mβ 2 q. Their (analytic)
solutions have the form g(t) = aeβt + be−βt . The I-state q(t) and motivation
p(t) are defined only for instants of I-time which satisfy the inequality
|βt|p ≤ rp . (76)
This is a consequence of properties of a p-adic exponential function, see
chapter 3, see also [163], [107], [111]. This condition can be considered as a
restriction for the magnitude of the I-force. If the coefficient of the interaction
|β|p ≤ rp , then the trajectory q(t) of the I-state is well defined for all t ∈ Zp .
Larger forces imply the restriction condition for I-time. Let |β|p = 1. If
p 6= 2 then (76) has the form t ∈ U1/p (0), i.e., t = α1 p + α2 p2 + · · · . Thus
the I-state q(t) of the I-transformer τ can be defined (observed) only for the
instants of time
t0 = 0, t1 = p, ..., tp−1 = (p − 1)p, ... .

114
If p = 2 then (76) has the form t ∈ U1/4 (0), i.e., and t = α2 22 + α3 23 + · · · .
Thus the I-state q(t) of τ can be defined (observed) only for the instants of
time
t0 = 0, t1 = 4, t2 = 8, ... .
2 2
Let f = −mβ 2 q, i.e., V (q) = mβ2 q and q̈ = −β 2 q. Here q(t) and p(t) have
the form q(t) = a cos βt + b sin βt. Here we also have the restriction relation
(76). As opposite to the real case the p-adic trigonometric functions are not
periodical. There is no analogue of oscillations for the I-process described
by an analogue of Hooke’s law.
Let us consider the solution of the Hamiltonian equations with the initial
conditions q(0) = 0 and p(0) = mβ: q(t) = sin βt, p(t) = mβ cos βt. We
have qp = (mβ/2) sin 2βt. By using the p-adic equality | sin a|p = |a|p we get
|qp|p = |mβ|p |βt|p . The relation (76) implies

|q|p |p|p ≤ |mβ|p rp . (77)

This is a restriction relation for the trajectory z(t) = (q(t), p(t)) in the phase
I-space (compare with [111]). Let β = 1/m. Then (77) gives |q|p |p|p ≤ rp .
If the motivation p is strong |p|p = 1, then q can be only of the form q =
α1 p + α2 p2 + · · · , p 6= 2 and q = α2 22 + α3 23 + · · · , p = 2. If the motivation
p is rather weak then the I-state q of an I-transformer can be arbitrary.
The restriction relation (77) is natural if we apply our information model
to describe psychological (social) behaviour of individuals. Strong psycho-
logical (social) motivations imply some restrictions for possible psychological
(social) states q. On the other hand, if motivations are rather weak an indi-
vidual can, in principle, arrive to any psychological (social) state.
We discuss the role of the I-mass in the restriction relation (77). There
the decrease of the I-mass implies more rigid restrictions for the possible
I-states (for the fixed magnitude of the motivation). If we return to the
psychological (social) applications we get that an individual (or a group of
individuals) with a small magnitude of I-mass and the strong motivations
will have quite restricted configuration of I-states.
The restriction relation (77) is an analogue of the Heisenberg uncertainty
relations in the ordinary quantum mechanics, [89]. However, we consider a
classical (i.e., not quantized) I-system. Therefore a classical I-system can
have behaviour that is similar to quantum behaviourAs we have already dis-
cussed, in chapter 5 the modern quantum formalism has two practically in-
dependent counterparts: (a) discreteness of observables - quanta; (b) Hilbert
space probabilistic calculus. At the moment we discuss similarities between
p-adic classical and quantum models related to (a) - quantum like discrete-
ness..

115
 We now study an P I-analogue of work and conservation laws for infor-
mation. Let f (x) = ∞ n
n=0 an x , an ∈ Qp , and let the series converge for
|x|p ≤ δ, δ = p±n , n = 0, 1, . . .. We define an integral of f by the formula
(see [187]):
Z b ∞
X an
f (x) dx = [an+1 − bn+1 ].
a n=0
n + 1

The series on the right-hand side converges for all |a|p , |b|p ≤ pδ . In particular,
Rx
we can find an antiderivative F of f by the formula F (x) = 0 f (x) dx.
Let f be an I-force which is described by the function f (x) which is
analytic
Rx for |x|p ≤ p. Then this force is potential with the I-potential V (x) =
0
f (x) dx.
Let γ = {q(t), |t|p ≤ λ} be an analytic curve in Qp . We define its length
element by ds = vdt, where v = q̇ is the I-velocity. By definition
Z Z t1
Wab = f ds = f (q(t))v(t) dt
γ(a,b) t0

where q(t0 ) = a and q(t1 ) = b. The quantity Wab is said to be the work done
by the external I-force f upon the I-transformer going from the point a to
the point b. Here a and b are points in the information space; so these are
tow I-states. By (56) we have
Z t1
m t1 d 2
Z
1 2
Wab = mv̇v dt = v dt = (p (b) − p2 (a)).
t0 2 t0 dt 2m

Thus the work done is equal to the change in the kinetic energy: Wab =
Tb − Ta . As the I-force
H f is potential then the work W done around a closed
orbit is zero: W = f ds = 0. Thus the work Wab does not depend on an
analytic trajectory γ(a, b). We also have:
Z Z
∂V d
Wab = − ds = − V (q(t)) dt = V (a) − V (b).
γ(a,b) ∂q γ(a,b) dt

Thus Tb − Ta = V (a) − V (b). We have obtained the energy conservation law

for an I-transformer:
Theorem 9.1. If the I-forces acting on an I-transformer are described by
analytical functions (in particular, they are potential), then the total energy
of the I-transformer, H = T + V , is conserved.
At the moment the situation with nonanalytic potential I-forces is not
clear. It may be that the energy conservation law is violated in the general
case.

116
36 Mechanics of a System of Information Trans-
formers, Constraints on Information Spaces
Let τ1 , . . . , τN be a system of I-transformers with I-masses,

m1 , . . . , mN ∈ Qp .

As in ordinary mechanics we must distinguish between the external I-forces

(e)
Fi acting on I-transformers due to sources outside the system and internal
forces Fji . As we have already discussed, I-times t1 , . . . , tN of τ1 , . . . , τN must
satisfy the consistency condition: t1 = t2 = · · · = tN = t. Thus the equation
of motion for the ith particle is to be written:
(e)
X
ṗi = Fi + Fji . (78)
j

For some I-systems we may obey an information analogue of Newton’s third

law (a law of information action and reaction): Fij = −Fji .
Set X
x= mi xi /M,
i
P
where M = mi . This point in the I-space is said to be the center of
information of the system. If the system satisfies Newton’s third law for
I-forces then we get the equation of motion:
X (e)
M ẍ = Fi = F (e) .
i

The center of information moves as if the total external I-force was act-
ing on the I-mass M of the system concentrated at the center of informa-
tion. We introduce the motivation P = M ẋ of the I-system. There is the
following conservation theorem for motions described by analytic functions
(qj (t))N
j=1 , t ∈ Qp :

Theorem 10.1. If the total external I-force is zero, the total motivation
of the I-system is conserved.
Example 10.1. (Social systems). We apply our I-model for describing a
society S which consists of individuals (or groups of individuals) τ1 , . . . , τN .
There exists the center of information of S, xS ∈ Qp . If S satisfies to New-
ton’s law of action-reaction for I-forces, then its evolution is determined by
the external I-forces. If this evolution is not pathological (so trajectories are
analytic), then the motivation of S is conserved. Of course, there might be

117
numerous pathological evolutions (for example, evolutions with zero motiva-
tion, PS = 0).
For analytic motions the work done by all I-forces in moving the system
from an initial configuration A = {ai = qi (t0 )} to a final configuration B =
{bi = qi (t1 )} is well defined:
XZ XZ
Wab = Fi dsi + Fji dsi
i γ(ai ,bi ) i6=j γ(a,b)

and Wab = TB − TA , where T = 12 i mi vi2 is the total kinetic I-energy of the

P
I-system. As usual
1 1X
T = M v2 + mi vi2 ,
2 2 i
where v is the velocity of the center of information and vi0 is the velocity of
τi with respect to the center of information.
In our model of social motion (Example 10.1) we can say that the total
kinetic energy of the society S is the sum of the kinetic energy of the center
of information of S and the kinetic energy of motions of individuals τj about
the center of information.
We now consider the case when all I-forces are (analytical) potential:
(e) ∂V
Fi = − ∂V i
∂xi
and Fji = − ∂xiji . To satisfy the law of action and reaction
we can choose Vij = Φij (xi − xj ) where Φij : Qp → Qp , Φij = Φji are
analytical functions. Then by repeating the considerations of the standard
mechanics
P over1 theP reals we obtain that WAB = −V (B) + V (A), where
V = i Vi + 2 i,j Vij is the total potential energy of the system of I-
transformers. Therefore
Theorem 10.2. The total I-energy H = T + V is conserved for every
I-system with (analytical) potential I-forces (such that Fij satisfy the law of
information action-reaction).
As in standard mechanics, constraints play the important role in I-mechanics.
The simplest constraints (holonomic) can be expressed as equations connect-
ing I-states of I-transformers τ1 , . . . , τN (or equations coupling different ideas
in an I-system):
f (q1 , . . . , qN , t) = 0.
Here f may be a function from QN p
+1
into Qp or a function from QN p
+1
into
R. The simplest constraints of the real type are:
(C1) |q1 − a|p = r, . . . , |qN − a|p = r, r > 0, a ∈ Qp , i.e., all I-transformers
have to move over the surface of the sphere Sr (a);
(C2) |q2 − q1 |p = r, . . . , |qN − q1 |p = r, i.e., there is the fixed I-transformer
τ1 such that all other I-transformers must move on the distance r from τ1 ;

118
(C3) We can also consider an ‘information rigid body’, i.e., a system of I-
transformers connected by constraints:
|qi − qj |p = rij .

Example 10.2. (Restricted mentality). In cognitive sciences constraint

(C1) can be used for the description of a ‘restricted mentality’. All ideas
q1 (t), ..., qN (t) of a cognitive system τ which are generated by the parallel
processes π1 , ..., πN belong to the restricted domain of ideas X = Sr (a).
In the case of holonomic constraints described by the system of analytical
functions: fj : QN p
+1
→ Qp , j = 1, . . . , K, i.e., fj (q1 , . . . , qN , t) = 0, we can
use the technique of the standard mechanics These methods may not be ap-
plied to constraints determined by real valued functions. However, in the lat-
ter case we need not eliminate these constraints. These constraints describe
open subsets of the configuration I-space QN p . We can choose such subsets
as new configuration I-spaces.. If the equations are independent then we can
introduce generalized I-coordinates ξ1 , . . . , ξN −K , and ql = ql (ξ1 , . . . , ξN −K , t)
l = 1, . . . , N , and ql (ξ, t) are analytical functions of ξ and t (see [109] for the
mathematical details).
Example 10.3. (Hidden basic ideas). If q(t) = (ql (t))N l=1 describes ideas
in the cognitive system at the instant t of I-time, then by resolving constraints
−K
on these ideas we can find independent ideas ξ(t) = (ξj (t))N j=1 which, in fact,
determine the I-state of a cognitive system.
Example 10.4. (Hidden leaders). If q(t) = (ql (t))N l=1 describes the
system S = (τ1 , . . . , τN ) of I-transformers then the existence of generalized
I-coordinates ξ can be interpreted as a possibility to reduce I-behaviour of
S to I-behaviour of the other system G = (g1 , . . . , gN −K ) of I-transformers.
We also remark that dynamics in I-space with constraints can be used
to describe aim-oriented thinking, see section 4, chapter 2. An aim can be
considered as a constraint on I-space which describes some surface. The
trajectory q(t) could not leave this surface in the process of motion. Of
course, an aim can be related not only to the I-state q of a cognitive system,
but also to his/her motivation p. Therefore, in general, we need to use aim-
constraints in mental phase-space.
As in the standard mechanics, we introduce generalized I-forces:
X ∂qi
Qj = Fi , (79)
i
∂ξj
where Fi is the total I-force acting to ith I-transformer, i.e.,
(a)
Fi = Fi + fi

119
 (a)
is the sum of applied force Fi and the force fi of constraints We can interpret
I-constraints as unknown I-forces..
In our theory generalized I-forces have the natural interpretation (com-
pare with the situation with generalized forces in the usual mechanics). As
we have noted, the existence of generalized I-coordinates which are obtained
from equations for constraints means that the initial system S = (τ1 , . . . , τN )
of I-transformers is controlled by other system G = (g1 , . . . , gN −K ) of I-
transformers. I-forces (79) are, in fact, reaction I-forces, i.e., the control of
G over S generates I-forces applied to elements of G. By repeating of the
usual computations we get the equations of motion:
d ∂T ∂T
( )− = Qj , j = 1, . . . , N − K. (80)
dt ∂ ξ˙j ∂ξj
∂V
If the I-forces Fi are potential with the analytical potential V , i.e., Fi = − ∂qi
,
∂V
then generalized I-forces are also potential: Qj = − ∂ξj . In this case the above
equation can be written in the form:
d ∂L ∂L
( )− = 0, (81)
dt ∂ ξ˙j ∂ξj

where L = T − V is the I-Lagrangian.

It is important that the equations (81) can be used to describe I-motions
in the presence of an I-potential V (q, q̇) which depends on (generalized) I-
velocities vi = q̇i . In this case
∂V d ∂V
Qj = − + ( ) (82)
∂qj dt ∂ q̇j

and L = T − V .
These velocity-dependent potentials may play an important role in I-
processes. In particular, there might be applications in such an exotic field
as anomalous phenomena. It is claimed (see, for example, [157]) that a
psychokinesis effect can be observed for some random physical processes and
it cannot be observed for deterministic processes. It might be tempting
to explain this phenomenon on the basis of the assumption that an I-field
generated in experiments on the psychokinesispsychokinesis corresponds to
a potential which depends on the I-velocity. Thus the corresponding I-force
is defined by (82). The I-velocity is higher for random processes. Therefore
the interaction is stronger for these processes.

120
37 Social and Anomalous Phenomena
1. Energy and information. In our model a transmission of information
is determined by the I-energy which is the sum of I-energy of motivations
and potential I-energy. In principle, this process need no physical energy.
Therefore, there might be transmissions of information which could not be
reduced to transmissions of physical energy. In this case we cannot observe
physical interactions, i.e., interactions in real space-time, between two I-
transformers, τ1 and τ2 . However, we might observe an I-interaction. In
particular, τ1 and τ2 can be individuals participating in psychological, [69],
[99], [139], [154], [157], [165]-[168], [200] ) or social experiments.
2. Distance and information. I-processes may evaluate in an I-space
which differs from the real space, absolute Newton space or spaces of general
relativity. Therefore the real physical distance between I-transformers does
not play a crucial role in processes of I-interactions. A cognitive system may
interact not only with other cognitive systems, but also with I-fields. In the
latter case an individual i might interact with some image which is located on
a large physical distance from i, see [69], [99], [139], [154], [157], [165]-[168],
[200].
3. Time and information. Dynamics of information is dynamics with
respect to I-time t. There may be a correspondence tphys = g(t) between real
time tphys ∈ R and I-time t ∈ Qp . This correspondence may not preserve
distances.
Let τ be an I-transformer having a continuous trajectory q(t). Small
variations of t, t0 = t + δt, imply small variations of q :
a0 = q(t0 ) = a + pδt, a = q(t). (83)
If (in some way) we find the internal time scale of τ, then it would be
possible to find (via (83) ) its I-state at the instant of time t0phys = g(t0 ). If
t0phys > tphys then such an I-measurement can be considered as a prediction
of future events; if t0phys < tphys then we have recalling. The relation (83)
gives only unsharp information. Thus such acts of recalling and predictions
may give a lot of ‘around’- information. It might be that such I-intersections
which violate the order structure of the real space might be identified with
some phenomena observed in parapsychology, see [69], [99], [123], [139], [154],
[157], [165]-[168], [200].
4. Motivation. A motion in the I-space depends, not only on the initial
I-state q0 , but also on the initial motivation p0 . Moreover, the Hamiltonian
structure of the equations of motion implies that the motivation p(t) plays
the important role in the process of the evolution. Thus I-dynamics is, in
fact, dynamics in phase I-space.

121
 5. Consistency for times. An I-interaction between I-transformers is
possible only if these I-transformers have consistent I-times. Therefore every
psychological or social experiment has to contain an element of learning for
I-transformers participating in the experiment. A physical interaction need
not be involved in such learning. This can be any exchange of information
between individuals (or a study of information about some individual).
6. Future and past. The consistency condition for I-times does not imply
such a condition for real times, because different I-transformers can have
different correspondence laws for I-time and real time. For example, let us
consider two I-transformers, τ1 and τ2 satisfying the consistency condition
for I-times, i.e., t1 = t2 = t. We assume that it is possible to transform
I-times of τ1 and τ2 to real times t1,phys = g1 (t1 ) and t2,phys = g2 (t2 ). Let
us also assume that τ1 and τ2 interact by the I-potential V (q1 − q2 ), i.e., at
the instant t of I-time the potential I-energy of this interaction equals to
V (q1 (t) − q2 (t)). If t1,phys = g1 (t) 6= t2,phys = g2 (t), then such an interaction
is nothing other than an interaction with the future or the past.
7. Social phenomena. By our model any social group G can be described
by a system τ1 , . . . , τN of coupled I-transformers. There exists an I-potential
V (q1 , . . . , qN ) which determines an I-interaction between members of G. For
example,
P democratic societies are characterized by uniform I-potentials V =
Φ(qi −qj ). Here a contribution into the potential I-energy does not depend
on an individual. On the other hand, hierarchic societies are characterized
by I-potentials of the form:
X X
V = A0 Φ(q0 , qj ) + A1 Φ(q1 , qj ) + · · ·
j6=0 j6=0,1

X X
+Ak Φ(qk , qj ) + B Φ(qi , qj ),
j6=0,...,k i,j6=0,...,k

where |A0 |p >> |A1 |p >> · · · >> |Ak |p >> |B|p . These potentials describe
the hierarchy τ0 → τ1 → · · · → τk → (τk+1 , . . . , τN ). The I-transformer τ0
can be a political, national or state leader or a God.
[Abstract Ultrametric Information Spaces] Abstract Ultrametric Infor-
mation Spaces
In this chapter we generalize our m-adic model of information space, I-
space. By using results of general topology and lattice theory we develop
I-models over various I-spaces having treelike structures corresponding to
ultrametric topologies.
This chapter is based on general theory of ultrametric spaces, [127], [57],
[72], [163]. First significant examples of such spaces were introduced in early

122
20th century in real analysis (Rene Baire, 1909) and topology (Felix Haus-
dorff, 1914)Hausdorff. In fact, the field of p-adic numbers Qp (introduced
by Kurt Hensel in 1894) with the metric ρp (x, y) = |x − y|p , where x → |x|p
is the valuation (an analog of the ordinary real absolute value) was the first
example of an ultrametric space. However, this fact was recognized only later
in the process of the development of general topology. General topology is
one of most abstract fields of modern mathematics. However, we present
topological results on the primary level. We hope that non-mathematicians
could follow to cognitive-topological considerations presented in this chapter.

38 Abstract Ultrametric Spaces

38.1 Thinking models over abstract ultrametric spaces
In chapter 2 we introduced the space of associations XA constructed over the
m-adic space of I-states. We remark that, in fact, each m-adic association
can be represented as a ball in the ultrametric space Zm and vice versa. By
using this correspondence between associations and balls we can generalize
cognitive constructions of chapter 2 to an arbitrary ultrametric space X.
Points x ∈ X are said to be I-states of a cognitive system (or general I-
system); balls

u = Ur (a) = {x ∈ X : ρ(a, x) ≤ r}, a ∈ X, r ≥ 0,

are said to be associations corresponding to families of I-states; collections

of associations are said to be ideas. The corresponding spaces are denoted
by symbols XI ≡ X, XA ≡ U (all balls), XID ≡ V (all collections of balls).
An actual radius of a ball Us (a) is the number r = sup ρ(a, x), x ∈ Us (a).
Clearly r ≤ s and r is uniquely defined. Further we consider only actual
radii.
If r = 0, then the ball Ur (a) = {a}. So the space of I-states XI is
embedded into the space of associations. The quantity o(u) = 1r , u = Ur (a),
is said to be the order of an association u. This definition differs from the
definition that was used for the space X = Zm in chapter 2. In chapter 2 we
used, in fact, the quantity o(u) = logm (1/r). In particular, I−states can be
considered as associations of the infinite order.
Let f : X → X maps a ball onto a ball: f (Ur (a)) = Us (b), b = f (a), r, s ≥
0. Such a map can be lifted from the space of I-states XI (≡ X) to the space
of associations XA (≡ U ) and then automatically to the space of ideas XID .
Dynamics in space X is described by iterations of dynamical system f and

123
its liftings to the spaces of balls U and collections of balls V, see (17), (18),
(19), chapter 2. The following question arises immediately:
‘Has such a model over an abstract ultrametric space any cognitive mean-
ing?’
The answer is ‘yes’. In fact, each abstract ultrametric space induces a
hierarchical system of trees, so called treelike lattice. Therefore dynamical
thinking in each ultrametric space can be represented as thinking on some
treelike lattice. Thus the study of models in abstract ultrametric spaces
supports our general ideology. Dynamics on m-adic hierarchical trees is just
the simplest thinking model. In general hierarchical chains of neurons can
have various tree-structures and generate various ultrametric spaces. After
my lectures at many Universities, I was often asked:
‘Do you claim that everybody has own natural number m such that
his/her brain has the m-adic structure?’
No, I do not claim this. But it seems that each individual has his/her
own ultrametric space (X, ρ) which describes the process of thinking. If two
individuals τ1 and τ2 have the same ultrametric I-space, then they can have
identical flows of mind. Another interesting mental property which can be
used for a classification of cognitive systems is related to the well known in
topology notion of isometry of metric spaces, [72].
We recall that a map j : X → Y, where (X, ρ) and (Y, ρ0 ) are metric
spaces, is called an isometry if
ρ0 (j(x1 ), j(x2 )) = ρ(x1 , x2 )
for all x1 , x2 ∈ X. We note that j maps balls of X onto balls of Y. Two metric
spaces are called isometric if there exists an isometry j from X onto Y. We
remark that an isometry is always one to one map.
Suppose that there are two cognitive systems τ1 and τ2 with ultrametric
spaces of I-states (X1 , ρ1 ) and (X2 , ρ2 ), respectively. If these spaces are iso-
metric, then there is one to one correspondence between I-states, associations
and ideas of τ1 and τ2 . Moreover, each thinking transformation f : X1 → X1
of τ1 induces the corresponding thinking transformation g : X2 → X2 of
τ2 : g(z) = j(f (j −1 z)), z ∈ X2 . By the isometric property of j this transfor-
mation can be lifted to the spaces of associations and ideas of a cognitive
system τ2 . We can say that τ1 and τ2 are isometric cognitive systems.isometric
cognitive systems

38.2 Treelike representation of abstract ultrametric spaces.

We need to recall a few definitions from the lattice theory, see [31], [66], [102]
for the details. A partial order ≤ on a set S is a relation between elements

124
of S such that:
(i) a ≤ a (reflexive);
(ii) a ≤ b, b ≤ a → a = b (anti-symmetric);
(iii) a ≤ b, b ≤ c → a ≤ c (transitive).
Let T be a subset of S. If g is a lower bound of T (so g ≤ t for all t ∈ T )
such that b ≤ g for every lower bound b of T, g is called the greatest lower
bound of T. In the similar way it is defined a least upper bound of T. If, for
any two elements x, y ∈ S, we can say that x ≤ y or y ≤ x, then the set S
is called linearly ordered. If, for a pair of elements a, b of a partially ordered
set, there is no x such that a < x < b, then it is said that the element a is
covered by b (or b covers a). An element a is called atom if the emptyset ∅
is the only element covered by a.
A lattice L is a partially ordered set in which every pair of elements a, b
has the greatest lower bound, a ∧ b, and a least upper bound, a ∨ b, within
this set.
A lattice L can be represented as a purely algebraic object, so called
(associative commutative) algebra - a set with operations of multiplication
and addition. A lattice L is an algebra with two additional postulates:
(π1 )a(a + b) = a, (π2 )a + ab = a.
Here we set ab = a ∧ b and a + b = a ∨ b.
We denote the set of atoms of a lattice L by the symbol A(L).
A lattice L is said to be treelike if the set L \ {0} is a tree, i.e., for any
element L ∈ L, L 6= 0, the set [L, →) of elements which are greater than or
equal to L, is linearly orderedIn the set-theoretical framework the element 0
is the empty subset ∅. In the algebraic framework the element 0 is zero of the
algebra.. Thus in general a lattice may be not linearly ordered, but, for each
element L, the set of larger elements is the linearly ordered set, the branch
of a tree.
We are going to establish the isomorphism between ultrametric I-space
(in fact, corresponding spaces of balls, associations) and some class of treelike
lattices. Here we shall use purely topological results,see [129]. We start with
the simple observation that the set L = U ∪ {0}, where U is the collection
of balls of an ultrametric space X and 0 = ∅, is a lattice with respect to the
order structure ≤ corresponding to the set-theoretic inclusion:
(a) for any two balls u1 , u2 there exists the greatest ball u = u1 ∧ u2 lying
in both u1 and u2 ;
(b) for any two balls u1 , u2 there exists the least ball u = u1 ∨u2 containing
both u1 and u2 ; it is either the greatest of two given balls (if one of them
contains the other) or the ball of radius r = ρ(x, y), where x ∈ u1 , y ∈ u2 ,
with a center located at any point of u1 ∪ u2 .

125
 This lattice is treelike. In fact, it is the consequence of the following
property of ultrametric balls: u ∩ v 6= ∅ → u ⊂ v or v ⊂ u. We remark that
the space of closed balls of a metric space which is not ultrametric does not
have such a tree structure. In particular, ordinary Euclidean space has no
such property, namely ‘Euclidean associations’ (closed balls of a Euclidean
space) could not be represented by a hierarchical tree. For example, let
X = R with ρ(x, y) = |x − y| and let L be the system of all segments [a,b]
(including X = R and ∅). Let L = [−1/2, 1/2] = U1/2 (0). The set [L, →) of
elements which are greater than or equal to L, is not linearly ordered: L ⊂
L0 = [−1/2, 32 ] = U1 ( 12 ) and L ⊂ L00 = [− 23 , 12 ] = U1 (−1/2). But L0 and L00 are
incompatible. Obviously the balls of zero radius (elements of X) are atoms.
The lattice L is atomic: for any ball u, u = sup{x : x ∈ u} = sup{x : x ≤ u}
(we use sup to denote the least upper bound of a set of elements).
A lattice E is said to be real graduated provided there is a real-valued
non-negative isotonic (see later) function r : E \ {0} → R ∪ {+∞}, satisfying
the following conditions:
(i) if x is an atom of E, then r(x) = 0;
(ii) if u1 < u2 , then r(u1 ) < r(u2 )
(iii) for any subset O of the set A(E) of atoms of E :
r(sup{x : x ∈ O}) = sup{r(x ∨ y) : x, y ∈ O}.

The real graduating function r gives the possibility to represent the order
hierarchy on a lattice by the standard order structure on the real line, see
(i), (ii). The (iii) is a technical condition. A map f is said to be isotonic if
u ≤ v → f (u) ≤ f (v).
The lattice L of balls is real graduated: r(u) =radius. The lattice L is
complete. We have the following proposition [129], p. 274.
Proposition 1.1. For any ultrametric space (X, ρ), the lattice L of its
balls is a complete atomic treelike real graduated lattice.
On the other hand, suppose we are given a complete, atomic, treelike real
graduated lattice E. As usual, denote by A(E) the set of its atoms. Let us
define a real valued non-negative function d(x, y) on A(E) × A(E) as follows:
d(x, y) = inf{r(L) : x, y < L} ≡ r(x ∨ y).
It is proven [129] that ρ is an ultrametric on A(E); so (A(E), d) is an
ultrametric space. It is proven [129] that if we, starting with an ultrametric
space (X, ρ) define the lattice L of its balls and then the ultrametric space
(A(L), d), then these two ultrametric spaces are isometric. Consequently,
if we starting with a complete, atomic, treelike lattice E define the ultra-
metric space of its atoms (A(E), d) and then the corresponding lattice L of

126
balls of the latter ultrametric space, then lattices E and L are isomorphic.
This purely mathematical result has the following important cognitive con-
sequence:
Theorem 1.1. Dynamical thinking in any ultrametric space can be rep-
resented as dynamical thinking on some treelike lattice and vice versa.
In some sense a lattice gives the spatial image of I-space (with its vertical
hierarchical structure). An ultrametric space represents the internal I-space
of a cognitive system.
S
Example 1.1 [66], [129]. Let X = xn = n/n + 1, n = 0, 1, 2, . . .} 1
and ρ(x, y) = max(x, y). This is an ultrametric space. The corresponding
lattice consists of ∅, elements of X (balls of zero radius), one ball of radius
1, L1 = U1 (1) = {xk ≤ 1}(= X ) which contains a chain of smaller balls
Ln/n+1 = Un/n+1 (n/n + 1) = {xk ≤ n/n + 1}. This is the main branch of this
tree. It is infinitely long. It has infinitely many subbranches of the length 2:
xn = n/n + 1 < Ln/n+1 , see Figure 7.1.

r r
1 q6 L1q q
16 Z
q 2 = U1 (0)
@ @
2 q L2 q @ U 1 (0)q @ q U 1 (1)
3
1 q L1 q 3 @ @ 2
q @q q @
@q 2
2 2 @
@ @
@ @ @
@q @
@ @q @q-
@ q q q q q q q q q q-
0 1 2 1X Z2
2 3

Figure 4: Lattice of balls for the ultrametric space X

Figure 5: Lattice of balls for the ultrametric space Z2

We remark that the lattice of p-adic balls reproduces geometrically the

structure of p-adic tree, see Figure 7.2 for p = 2. However, the metric
structures of the tree on Figure 7.2 and 2-adic tree for I-states, see Fig-
ure 1.1 (chapter 1), are essentially different. On the lattice of balls we have
ρ(u, u0 ) = 0 if u and u0 lye on the same line and r(u) < r(u0 ) (i.e., u ⊂ u0 ).
Moreover, on Figure 7.2 points are vertexes of the 2-adic tree and, for the
original 2-adic tree, points are branches of this tree (on Figure 7.2 these
branches are represented by points on the horizontal level r = 0.)

127
39 Hierarchy of Associations
The Hausdorff pseudometric ρ(A, B) = supa∈A inf b∈B ρ(a, b) can be modified
to obtain a metric on the space of balls U of an ultrametric space X, see [72],
by setting
d(A, B) = max[ρ(A, B), ρ(B, A)].
The space of balls U becomes an ultrametric space (U, d). The correspond-
ing space of balls of this space U (1) can be again represented as a treelike
lattice. Therefore, it would be better to define ideas not as arbitrary subsets
of the space of associations XA ≡ U, but as elements of the space of balls
U (1) . Here only special collections of associations are considered as cognitive
structures, ideas.
In all following sections of this chapter we define ideas as balls of balls of
an ultrametric space (with respect to Hausdorff ’s metric).
We underline that such a definition of an idea does not coincide with the
definition given in chapter 2.
We can again introduce the Hausdorff metric on U (1) and repeat the
procedure. In principle, we obtain an infinite sequence of lattices

L ⊂ L(1) ⊂ L(2) ⊂ . . . .

Here L = U is the lattice of balls of an ultrametric space X in that the

primary I-dynamics is performed, L(1) = U (1) is the lattice of balls of U,
(so U = A(L(1) ) is the set of atoms of L(1) ) and so on. By forming hier-
archical (balls-like) structures of I-states we create associations; by forming
hierarchical (balls-like) structures of associations we obtain ideas which we
called ideas of order 1. In fact, these ideas are nothing than associations of
associations. We can continue this process infinitely long.
Conjecture 1. Cognitive systems (at least some of them) are able to
operate simultaneously on all levels of the infinite cognitive hierarchy. In
such a case a cognitive system would be able to operate with an infinite
volume of information on the basis of a finite physical system.
Conjecture 2. Consciousness is created by this infinite volume of infor-
mation which is concentrated in a finite domain of physical space.

40 Topology and Materialism

There exists numerous topological results on embedding of ultrametric spaces
in Euclidean spaces and arbitrary metric spaces into ultrametric spaces, see
[72]. In our cognitive model of dynamical thinking in ultrametric spaces

128
some of these topological results may have a cognitive interpretation. We
start with Theorem of A. Lemin [130]:
Theorem 3.1. Every ultrametric space consisting of n + 1 points can
be isometrically embedded in the n-dimensional Euclidean space Rn . No ul-
trametric space consisting of n + 1 points can be isometrically embedded in
Rk , k < n.
It seems that this theorem might be interpreted as the evidence of im-
possibility of ‘spatial localization of mind’ in brain.
The standard physical space in that we consider ‘physical’ brain is R3 .
So, in fact, in this space we could isometrically embed ‘mental’ brain of a
cognitive system τ having only 4 different I-states. As human brain might
represent a huge number N of I-states, it could be described only as a ‘phys-
ical body’ in the Euclidean space E N of a huge dimension. Therefore, each
attempt to produce a 3-dimensional map of ‘mental brain’ would imply non-
isometric embedding: some similar (close) minds should be represented by
far configurations in R3 and some essentially different minds should be repre-
sented by similar configurations in R3 . It is interesting that our ideal I-space
Zp can be isometrically imbedded in a separable Hilbert space, see [130].
Therefore, the p-adic thinking models which are developed in this book,
in fact, describe (rather special) dynamics on a simplexp in a Hilbert space.
This simplex is located on the sphere of radius r(p) = p/ 2(p2 + p + 1), see
[130].
Moreover each ultrametric space can be isometrically embedded in some
generalization of a Hilbert space [130]. Therefore, Hilbert space and its
generalizations are sufficient for a ‘spatial’ realization of all possible I-spaces.

41 Existence of Universal Mental Space

In this section we discuss a possible cognitive interpretation of Lemin-Schepin
theorem (private communication of A. Lemin) on ultrametrizationultrametrization
of an arbitrary metric space (X, ρ) :
Theorem 4.1. There are an ultrametric space (X̃, ρ̃) and non-expanding
map u from X onto X̃,
u : X → X̃, u(X) = X̃ , (84)
such that, for any non-expending map f : (X, ρ) → (Y, d), where (Y, d) is an
ultrametric space, there is a unique non-expending map uf : (X̃, ρ̃) → (Y, d)
such that f (x) = uf (u(x)), x ∈ X.
We recall that a map f is said to be nonexpanding if
d(f (x1 ), f (x2 )) ≤ ρ(x1 , x2 ).

129
We consider ultrametrization from the cognitive viewpoint. This fact implies
that any physical metric space (X, ρ) could be ‘understood’ by a cognitive
system τ having the ultrametric space (X̃, ρ̃) as a space of I-states. The map
(84) gives the representation of a physical space X in a mental space X̃. For
a geometric figure O ⊂ X, the u(O) gives the mental representation of this
figure.
So physical space (X, ρ) can be represented by a treelike lattice and phys-
ical dynamics in X can be translated into dynamics on such a lattice. In
particular a neural network of neurons in our brain is nothing than uni-
formization of the Euclidean space R3 .
What is about the Minkovski’s (pseudo-Euclidean) space R3,1 ? It seems
that (unfortunately) networks used in our brain could not produce cognitive
images of objects in R3,1 . We have the following topological evidence of this
[130]:
Theorem 4.2. If distances between points a0 , . . . , an satisfy the strong
triangle inequality, then the subspace spanned by them is Euclidean.
Two metric spaces (X1 , ρ1 ) and (X2 , ρ2 ) are said to be u-equivalent if
their ultrametrizations (X̃1 , ρ̃1 ) and (X̃2 , ρ̃2 ) are isometric. As we have al-
ready mentioned, isometric ultrametric spaces produce equivalent thinking
processes (in spaces of I-states as well as associations and ideas).
Therefore, we obtain a possibility to classify general physical metric spaces
via their cognitive representations, ultrametrizations (or the corresponding
lattice representations). Two u-equivalent physical spaces would produce
‘equivalent’ cognitive civilizations.
We now discuss the second part of the ultrametrization theorem, namely
the representation of any non-expending map f : X → Y as f = uf u, where
u : X → X̃ is an ultrametrization map and uf : X̃ → Y is a non-expending
map between ultrametric spaces.
Let us consider the following situation. There is a physical metric space
(X, ρ) and there are two cognitive systems τ1 and τ2 with ultrametric spaces
(X1 , ρ1 ) and (X2 , ρ2 ) of I-states. Suppose that τ1 and τ2 developed the abil-
ity of a mental representation of the physical space (X, ρ) : there are two
non-expending maps f1 : X → X1 and f2 : X → X2 which give such men-
tal representations of X by τ1 and τ2 , respectively. The property that f1
and f2 are non-expending is very important for the right orientation in the
environment: if points x1 and x2 are close in physical space X, then their
(j) (j)
mental images α1 = fj (x1 ) and α2 = fj (x2 ) must be close in the I-space
Xj (j = 1, 2). This produces the image of continuity of the physical space X
(in the corresponding I-space). The existence of the ultrametrization (X̃, ρ̃)
of the physical space (X, ρ) can be interpreted as the topological evidence of
the possibility to create a universal cognitive system τuniv (for the physical

130
space X) that could ‘understand’ all cognitive representations of X. Let x
be a point of X and α(j) = fj (x), j = 1, 2, and α = u(x) be mental repre-
sentations of this physical point by τj , j = 1, 2 and τ, respectively (in spaces
Xj , j = 1, 2, and X̃, respectively). Then α(j) = ufj (α), j = 1, 2, where
ufj : X̃ → Xj are maps corresponding to fj : X → Xj (via Lemin-Schepin’s
theorem). So mental images α(j) of Tj are representations of the same mental
image α of τ. As maps ufj are non-expending, a small variation of an image
α → δ + α produces a small variation of an image α(j) → α(j) + δα(j) . Hence
τj could ‘understand’ τ.
In the case of an expending map a small variation of τ ’s mind may produce
an essential variation of its image in the I-space Xj of τj . So, if ufj : X̃ → Xj
is interpreted as a communication channel between cognitive systems τ and
τj , then a non-expending property of this channel gives the possibility to
escape misunderstandings.

42 Towers of Associations
Let (X, ρ) be an ultrametric space; let (U, ρ) be the space of closed balls of
(X, ρ) endowed with Hausdorff’s metric and let (V, ρ) be the space of closed
balls of (U, ρ) endowed with Hausdorff’s metric. We denote balls in (X, ρ) by
symbols Ur (a), where r is the actual radius; balls in (U, ρ) by symbols Vd (u),
where u = Ur (a) ∈ U and d is the actual radius. The following two Lemmas
describe the geometry of balls in the space U (so balls of balls). The reader
who is not interested in mathematical details can just pay attention to the
formulation of Corollary 5.1 (starlike structure of balls in the space V ).
Lemma 5.1. Vr (Ur (a)) = {Us (b) : b ∈ Ur (a), s ≤ r}.
Proof. 1. We shall prove that each ball Us (b), b ∈ Ur (a), s < r, belongs
to the ball Vr (Ur (a)). By definition r = sup{ρ(x, b) : x ∈ Ur (b) = Ur (a)}.
There exists xn ∈ Ur (b) such that r = ρ(xn , b) + n , n → 0.
As s < r, ρ(xn , b) > s, for n ≥ N, for a sufficiently large N. Hence,
ρ(xn , y) = ρ(xn , b) for all y ∈ Us (b). Thus inf y∈Us (b) ρ(xn , y) = r − n →
r, n → ∞. Therefore we obtain:

ρ(Us (b), Ur (a)) = r, s < r, b ∈ Ur (a). (85)

2. We demonstrate that if, for some ball Us (b), we have ρ(Us (b), Ur (a)) ≤
r, then Us (b) ⊂ Ur (a) (i.e., s ≤ r and b ∈ Ur (a)) :
2a. Let s = r : r ≥ ρ(Ur (b), Ur (a)) ≥ inf y∈Ur (a) ρ(b, y). But

ρ(b, a) ≤ max[ρ(b, y), ρ(y, a)], y ∈ Ur (a).

131
Hence ρ(b, a) ≤ r and Ur (a) = Ur (b).
2b. Let s > r. Suppose b 6∈ Ur (a), i.e., ρ(b, a) > r. Hence, ρ(b, y) =
ρ(b, a) > r for all y ∈ Ur (a). Thus we get ρ(Us (b), Ur (a)) > r. Suppose
b ∈ Ur (a). As s > r, we get Ur (a) ⊂ Us (b). Hence ρ(Us (b), Ur (a)) = s > r.
Therefore it is impossible that s > r.
2c. Let s < r. Suppose b 6∈ Ur (a), i.e., ρ(b, a) > r. Here

ρ(Us (b), Ur (a)) ≥ sup inf ρ(x, y) ≥

x∈Ur (a) y∈Us (b)

inf ρ(a, y) = ρ(a, b) > r.

y∈Us (b)

Suppose b ∈ Ur (a). Here Us (b) ⊂ Ur (a).

Corollary 5.1. The ball Vr (Ur (a)) has a star structure:
[
Vr (Ur (a)) = Ur (a) Sr (Ur (a)).
Center of this star is the ball Ur (a), rays are directed from center to balls
Us (b) ⊂ Ur (a) which are located on the sphere of radius r around center. We
note that the ball Vr (Ur (a)) contains only one balls of radius r. Therefore
each ball in (V, ρ) can be uniquely represented as a star.
Corollary 5.2. Let s < r. Then Vs (Ur (a)) = {Ur (a)}.
So there are no nontrivial balls of radius s < r with center at Ur (a).
Corollary 5.3. Let s > r. Then Vs (Ur (a)) = Vs (Us (a)).
We recall that by definition U0 (a) = {a}, a single point (so X is embedded
into U ).
Lemma 5.2. For any a ∈ X, Vr (a) ≡ Vr (U0 (a)) = Vr (Ur (a)) = {Us (b) :
b ∈ Ur (a), s ≤ r.}
Proof. 1). Let b ∈ Ur (a), s ≤ r. Then

ρ(Us (b), a) = sup ρ(y, a) ≤ sup max(ρ(y, b), ρ(b, a)) ≤ r.

y∈Us (b) y∈Us (b)

2. Let, for some ball Us (b), we have ρ(Us (b), a) = supy∈Us (b) ρ(y, a) ≤ r.
Then ρ(b, a) ≤ r, so b ∈ Ur (a), and
s = supy∈Us (b) ρ(y, b) ≤ supy∈Us (b) max(ρ(y, a), ρ(a, b)) ≤ r.
Moreover, we can calculate ρ(Us (b), a) precisely:
a) Let a ∈ Us (b). So Us (b) ≡ Us (a). Here ρ(Us (a), a) = supy∈Us (a) ρ(y, a) =
s.
b) Let a 6∈ Us (b). So ρ(a, b) = d > s. Here ρ(y, a) = max(ρ(y, b), ρ(b, a)) =
d for all y ∈ Us (b).
Thus all nontrivial balls in the space U (elements of the space V ) are
given by balls Vr (Ur (a)).

132
43 Infinite Information Towers
Lemmas 5.1,5.2 give the possibility to find the cardinality of I-spaces on
different floors of the mental tower. We set X (0) = X, X (1) = U, X (2) = V and
so on; denote by NK the cardinality (number of elements) of the space X (K) .
We have: N1 = N0 +N+ , where N+ is the number of balls Ur (a), r > 0. In the
same way N2 = N1 +N+ , because each ball Vr (u), u ∈ X (1) , r > 0, corresponds
to the ball of the same radius in X (1) , u = Ur (a). So N2 = N0 + 2N+ and by
induction we obtain that NK = N0 + KN+ . Therefore the I-power of mental
processing increases linearly in the vertical direction.
Example 6.1. Let X = Zm and ρ = ρm . Here the space of balls U has
the following branches:

U1 (0) ⊃ U 1 (0), U 1 (1), . . . , U 1 (m − 1);

m m m

U 1 (0) ⊃ U 1 (0), U 1 (m), . . . , U 1 (m2 − m)), . . . .

m m2 m2 m2

Balls of V are of the form: V1 (U1 (0)) = U, this is the star with center at
U1 (0) and rays directed to points lying on the sphere S1 (U1 (0)), namely all
elements of U besides of U1 (0); V 1 (U1 (0)) = {U1 (0)};
m
V1/m (U1/m (0)) = {U1/m (0), U1/ml (a), a = (a0 , . . . , al−1 ), a0 = 0, l = 2, . . . , ∞}.
This is the star with center at U1/m (0) and rays directed to points lying
on the sphere S 1 (U 1 (0)), namely all sub-balls of the ball U 1 (0).
m m m

We calculate the I-capacity of an m-adic cognitive system which operates

with I-strings of the length M. Here N0 = mM . We have for the cardinality
N1 of the set U : N1 = N0 + N+ , where N+ is the number of balls with
r > 0. It is easy to see that N+ = 1 + m + . . . + mM −1 = mM − 1/m − 1
(we remark that balls of the radius r = 1/mM have the actual radius r = 0).
Thus, for the cardinality NK of the I-space of kth order (associations of ...
of associations), we have

NK = mM + K(mM − 1/m − 1).

Therefore the I-capacity of the m-adic coding system increases exponen-

tially with respect to the length M of hierarchical chains and linearly with
respect to mental levels. The total I-capacity of the mental tower having n
floors and the structure of the m-adic tree with branches of the length M on
the ground floor is equal to

N (m, M, n) = nmM + 1/2n(n + 1)(mM − 1/m − 1).

133
 If M = const and n → ∞, then N ≈ Cn2 , c > 0. We discovered the surprising
feature of cognitive systems which are able to create mental towers with
vertical hierarchy, namely producing new trees of trees.
By using a physical system, ‘brain’, of a finite I-capacity (for example,
an ensemble of hierarchical chains of neurons of the length M < ∞) such
a cognitive system is able to create the I-capacity N ≈ cn2 → ∞, when the
number of mental levels for I-processing n → ∞.
We can even speculate that the phenomenon of consciousness might be
related to such an ability to create mental towers with horizontal I-hierarchy.
It might be that the parameter n describes a level of consciousness of a cogni-
tive system. For example, human beings are characterized by the inequality
n ≥ nHB , where nHB is a constant which could be in principle found exper-
imentally. Different individuals may have different values of n ≥ nHB . In
principle, it may be possible to find a probability distribution of levels of
consciousness.
We remark that there are no reasons to reject the possibility of the cre-
ation of cognitive systems having infinitely high mental towers, n = ∞,
based on a finite physical system, ‘brain’. However, it is not clear how such
a cognitive system could evolve from cognitive systems with finite n.
Let X = {1, 2, 3, . . . N } be the set of the first N natural numbers endowed
with the ultrametric ρ(x, y) = max(x, y), x 6= y. Here U consists of the balls
un = Un (n) = {j : j ≤ n}, balls of zero radius, namely elements of X, and
the whole space X = UN (j) (for any j).
Geometrically U can be represented by the following tree:
uN r
B
B
uN −1 r B
B B
B
u2 r
B
B B
B B B
B
u1 = 1 r BrN
B B
B B
B B
BrN −1
B B
B
B
Br2
B

Figure 6: Tree-representation of the maxmetric space for N natural numbers

The main branch uN ⊃ uN −1 ⊃ . . . ⊃ u1 ⊃ ∅ produces subbranches:

uN ⊃ {N }, . . . , u1 ⊃ {1}. We note that the cardinality of X, |X| = N, and

134
the cardinality of U, |U | = 2N − 1 (the ball U1 (1) = {1} ≡ U0 (1), therefore
only points 2, 3, . . . , N produce balls of r > 0).

135
 We can visualize elements of the space of associations U corresponding to
the three-point metric space X = {1, 2, 3} in the following way (for example,
a square is associated with a triangle via the image u3 ) :

1 2 3@
@
@

@
@
@

u2 u3

Figure 7: Images of associations

We go to the next floor of the mental tower. We have v1 = V1 (u1 ) =

{u1 } = {1}, v2 = V2 (u2 ) = {u2 , 1, 2},
v3 = V3 (u3 ) = {u3 , u2 , 1, 2, 3}, . . . , vN = VN (uN ) = {uN , . . . , N }.

 
1 u2

'$
@
@
u3
&%
 @
@
2 3
 

Figure 8: Idea V3 (u3 )

The tree of the space of ideas, see Figure 7.6 is essentially more complex
then the tree of the space of associations. We now consider the third floor of
the mental tower (with X as the ground floor), the space of ‘super-ideas’, see
Figure 7.7. The corresponding space W = {1, 2, . . . , N, u2 , . . . , uN , v2 , . . . , vN , w2 =
W2 (v2 ) = {1, 2, u2 , v2 }, . . . , wN =
WN (vN ) = {1, . . . , N, u2 , . . . , uN , v2 , . . . , vN }}.
As usual, elements wj has a star-structure: vj is center and elements

1, . . . , j, u2 , . . . , uj , v2 , . . . , vj−1

are located on the sphere Sj (vj ).

136
 [Pathway Representation of Cognitive Information] Pathway Representation
of Cognitive Information
One held that psychological functions such as language or memory could
never be traced to a particular region of the brain. If one had to accept,
reluctantly, that the brain did produce the mind, it did so as a whole and
not as a collection of parts with special functions. The other camp held that,
on the contrary, the brain did have specialized parts and those parts generate
separate mind functions. A.R. Damasio.
In this chapter we develop a cognitive information model based on the
representation of cognitive information by centered neural pathways. This
model is a natural generalization of the model based on the representation
of cognitive information by hierarchical chains of neurons (HCN ) which is
developed in chapter 2, see also chapter 5. We propose a cognitive model in
which Mental States are described as probability distributions of information
on the space of all centered neural pathways representing a special psycho-
logical function (processor). Each psychological function is based on its own
hierarchical tree of centered neural pathways, a cognitive tree. Hence firings
of individual neurons do not have any cognitive meaning in this model, see
chapter 2. Only hierarchically organized pathways firings give contributions
to a Mental State. The process of thinking is described by evolutionary equa-
tions in the space of probability distributions representing Mental States. As
in all previous chapters, the ultrametric p-adic geometry (which differs essen-
tially from the standard Euclidean geometry) on the I-space plays the crucial
role in our considerations. We also discuss the role of body, sensations, and
emotions (compare with A. Damasio [52]) in the process of neural pathway
thinking.
Since [119] we have developed a model of the process of thinking based
on the neural pathway representation of cognitive information, the Neural
Pathway Approach,neural pathway approach see chapter 2. In our model the
elementary unit of cognitive information is given not by the frequency of firing
of an individual neuron, but by the string of firings of neurons throughout a
pathway of neurons. Such a string of firings throughout a pathway we call
an I-state. As always, we shall use the symbol XI to denote the space of
I-states – I-space.
The crucial point of our investigation is understanding that each psy-
chological function (processor) is based on a tree of neural pathways, the
cognitive tree,cognitive tree which is centred with respect to one fixed neuron
S. Such a centring determines a hierarchical structurehierarchical structure
on a cognitive tree and on the corresponding space of I-states. This hier-
archical structure induces the p-adic ultrametricultrametric geometry on the
I-space XI .

137
 Of course, such a model with one neuron centring of a psychological
function is over-simplified. Complex psychological functions should be based
on a few cognitive trees centered with respect to an ensemble of neurons.
We remark that in chapter 2 we have not paid attention to a tree structure
of neural pathways. In fact, we directly jump to the I-space XI of I-states
produced by centered neural pathways. In principle such an I-space arises
even for one centered pathway — hierarchical chain of neurons (HCN , see
chapter 2):
N = (n0 , n1 , ..., nM , ....). (86)
Of course, we pointed out that ensembles of HCN s play an important role
in the representation of cognitive information. We recall the important fact
that was discussed in chapter 2: a fixed I-state x ∈ XI can be represented
by an ensemble of neural pathways. This fact is the cornerstone of a prob-
abilistic cognitive model that will be developed in the present chapter. The
important difference between models presented in chapter 2 and this one is
that in the former model we considered not all centered neural pathways, but
only HCN s starting with the centring neuron, see (86) where S = n0 .Thus
when we speak about centring this is not centring with respect to Euclidean
geometry. In this context centring has the meaning of importance. We can
compare this with the notion of a political center. Such a center (e.g., the
capital of a country) need not be a geographical center. This is the very
restricted class of centered neural pathways. Here the central neuron S = n0
does not have any input. The whole HCN is ‘ruled’ by the S. In the present
chapter we consider the general case: centered neural pathways with neurons
producing input for the S :

N = (n−j , ..., n−1 , S = n0 , n1 , ..., nM , ....). (87)

In chapter 2 the process of thinking was reduced to operating with I-states

as well as higher order information structures composed of I-states, namely,
associations and ideas. The operating had the meaning of functioning of
a dynamical system on the space of I-states – strings of firings of neurons
throughout pathways. We considered a dynamical system, a feedback pro-
cess, in the mental space XI . Mathematically such a dynamical system was
described by a map f : XI → XI that mapped strings of firings throughout
pathways into strings of firings throughout pathways. In such a simplest
model the process of thinking was described by a mathematical law:

xn+1 = f (xn ), (88)

where x belongs to the I-space XI . In fact, this approach to the process

of thinking is a natural extension of the Neural Dynamical Approach, see,

138
e.g., [11], [12], [70], [93]. The main distinguishing feature of our approach is
that instead of studying dynamics of firings of individual neurons we study
dynamics of firings of whole pathways. The string of firings is considered as
the elementary unit of mental information. Our approach can be called the
Neural Pathway Dynamical Approach.
In this chapter we develop the Neural Pathway Approach. The main
problem of the dynamical model studied in chapter 2 was the absence of a
description of the body-mind relation and its role in the process of think-
ing. In this chapter we provide such a description by identifying the mental
statemental state of a cognitive system with the probability (intensivity) of
the representation of an I-state by neural pathways. In the present model the
mental process is described as a stochastic process performing ‘body→mind’
relation. The evolution of the mental state — the probability distribution of
this process — can be described (at least for simple psychological functions)
as diffusion on an ultrametric p-adic tree: thinking as ultrametric diffusion.
Psychological, neurophysiological, cognitive, and philosophical consequences
of our thinking model (based on the probability distribution on the space of
neural pathways) will be discussed in the next section.

44 Model: Thinking on a Cognitive Tree

As has already been mentioned in chapter 2, one of the strong sides of the
Neural Pathway Approach is a new viewpoint on the problem of localization
of psychological functions. Since the elementary cognitive unit is represented
by a pathway and a pathway can go through various domains of brain and
body, there is no localization of mental functions in the Euclidean geometry
of the physical space (which is typically used to describe physical brain and
body). On the other hand, there is localization in the ultrametic space of all
pathways. In fact, this is a kind of hierarchical localization — compare with
A. Damasio:
“What determines the contribution of a given brain unit to the operation of the
system to which it belongs is not just the structure of the unit but also its place
in the system. ... The mind results from the operation of each of the separate
components, and from the concerted operation of the multiple systems constituted
by those separate components", p. 15, [52].
In our model there is even no place for ‘separate components’; everything
is unified from the beginning as a consequence of the pathway representation
of cognitive information.
We have to distinguish the space Π of all pathways, chains of neurons, in

139
the physical brain and body and the space XI of all possible I-states that
can be produced by elements of Π. In principle, a few distinct elements of
Π, pathways, can produce (at some instant of time) the same element of XI ,
an I-state. Moreover, it should be the case, since it is very dangerous for
a cognitive system to base the information representation of an important
I-state by using a single neural pathway. Thus (at least some) I-states
should be represented by (perhaps very large) ensembles of neural pathways.
In fact, this multiplicity of the neural pathway representation of I-states
might be the main fundamental feature of the process of thinking, see further
considerations on the probabilistic structure of mental states.
And this is not the end of the psychological function localization story
in the Neural Pathway Approach. The crucial point of our considerations is
that:
The most natural (I could say beautiful) mathematical model, so called p-
adic geometry on the I-space XI , is obtained under the assumption that each
pathway contains a Central NeuronCentral Neuron – S. Roughly speaking, S
collects information from the preceding part of the pathway and distributes
this information throughout the following part of the pathway. By choosing
the central neuron we obtain a hierarchical structure on the space XI . The
corresponding geometry on XI is so called ultrametric space geometry.
We do not yet have neurophysiological and psychological confirmations
that thinking is based on the ultrametric geometry. However, the mathemat-
ical beauty of the model is a strong argument (as it very often is in physics)
in favour of the ultrametric p-adic model in the Neural Pathway Approach.

Remark 1.1. By choosing the central neuron S we choose the center

of a coordinate system in the (p-adic) ultrametric geometry. By choosing
a system of coordinates we choose a psychological function. It is impor-
tant to underline that we do not claim that there exists a kind of absolute
central neuron or a group of neurons that ‘rule’ all mental processes. Our
geometric model of mental processing is not similar to the model of physical
processes based on Newtonian absolute space. Our model is closer to models
of relativity theory.
We now turn back to the psychological function localization story. There
is no (Euclidean) localization of a psychological function. However, there is
partial localization related to the central neuron S of the tree of pathways
representing a psychological function.
In chapter 2 we have studied the Neural Pathway Dynamical Model for
the dynamics of I-states produced by one fixed central pathway. Of course,
we understood that the model was oversimplifiedNevertheless, even such a

140
model has incredibly rich cognitive and mathematical structures.. First of
all from neurophysiological point of view it would be natural to suppose that
a psychological function (in advanced cognitive systems) is based not on a
single centered pathway, but on a system of such pathways. In the simplest
case all these pathways are centered with respect to the same central neuron
S. Therefore it would be useful to consider a tree of pathways, see, e.g.,
Figure 8.2.
Firings of neurons throughout pathways of the cognitive tree produce
elementary I-states involved in the realization of a psychological function.
We denote a psychological function by the symbol f and a cognitive tree
used by the f by the symbol Πf . How should we represent mathematically
the functioning of f ? There are various levels of the description. At the basic
level we should provide the description of ‘body→mind’ correspondence. This
correspondence is described by a function

ϕ : Πf → XI

that maps neural pathways into I-states represented by these pathways: z ∈

Πf → x = ϕ(z) ∈ XI . We call the map ϕ(z) the body → mind fieldOf course,
φ depends on the psychological function f, namely, φ = φf ..
The psychological function f performs the evolution of the field ϕ. Start-
ing with the initial field ϕ0 (z), f produces a time-dependent field ϕ(t, z).
We have to consider the very important problem of the interpretation
of the evolution parameter, the ‘time’, t. We shall discuss this problem in
section 6. At the moment we restrict ourselves to considering the discrete
time evolution: t = tn = 0, 1, 2, . . .. By taking into account the process of
wholeness of thinking we describe the functioning of f by an integral operator
with the kernel K(z, y) :
Z
ϕ(tn+1 , z) = K(z, y)ϕ(tn , y)dy, (89)
Πf

where the integration is performed over the cognitive tree, Πf . We notice

that neither the space of pathways Πf nor the space of I-states XI have the
Euclidean geometry. In particular, we cannot use the ordinary real analysis
to describe this model mathematically. We must use the ultrametric analysis.
The form of the kernel K(z, y) is determined by the psychological func-
tion f : K = Kf . We notice that mental evolution (89) is represented by
a linear integral operator in the space of body→mind fields. In principle,
we can consider more general, nonlinear models. However, the model with
summation over the whole tree with a weight function K(z, y) looks very
natural.

141
 Thus we propose the following model of thinking:
Each psychological function f is based on a tree of neural pathways,
a cognitive tree Πf , see, e.g., Figure 8.2. The cognitive tree Πf , has the
hierarchical structure corresponding to the existence of the central neuron,
S, of this tree. An elementary unit of information is given by the string of
firings of neurons throughout a pathway, a branch of the tree. There can be
proposed various neural pathway coding systems based on strings of firings:
2-adic coding (‘yes–no’ coding). For each instant of time t we assign
to each neuron in a pathway — 1, if a neuron is firing, and 0 otherwise.
Here mathematically an I-state is represented by a sequence of zeros and
ones. Each sequence is centered with respect to the position corresponding
to firings of the central neuron S.
Let us consider the geometric structure of the I-space XI corresponding
to the simplest cognitive tree of centered pathways. There are no input neural
pathways going into the central neuron S. Here each I-state belonging to the
corresponding I-space can be coded by a sequence of zeros and ones. The
first digit in a sequence gives the state, 0 or 1, of the central (in this case
the starting) neuron S. Hierarchy on the I-space is based on the exceptional
role that is played by the central neuron. This hierarchy induces the 2-
adic ultrametric topology on the I-space. As we already know, this space is
nothing other than the ring of 2-adic integers Z2 . Here the distance (induced
by 2-adic valuation) between two I-states x and y is small if they have a very
long common root starting with the S. Thus in our model:
Neurons located on different pathways at large distances from the central
neuron can produce very close I-states!
p-adic coding. General p-adic coding (where p ≥ 2 is a natural number)
may be induced by the frequency coding. We assign to each neuron in a
pathway the frequency of firings. In the mathematical model it is convenient
to consider a discrete set of frequencies: 0, 1, .., p − 1, where p is some natural
number. Here frequency is the number of output spikes produced by a neuron
during some unit period of time, e.g., one second. Thus mathematically
an I-state is represented by a sequence of numbers belonging to the set
{0, 1, . . . , p − 1}. Information is not homogeneously distributed throughout
such sequences. The presence of the central neuron S in the cognitive tree
Πf induces a hierarchical structure for elements of an information sequence.
Here the cognitive tree Πf in general produces the I-space XI = Qp .
We underline that the system of coding and not the topological structure
of a cognitive tree determine the structure of the corresponding I-space, see
also section 6. Totally different cognitive trees Πf can produce the same

142
mental tree XI = Qp . For example, let us consider 2-adic coding. The trees
in figures 8.3 and 8.4 produce the same, 2-adic, I-space.
For each point z (pathway) of the cognitive tree we define the I-state
ϕ(z) produced by this pathway. There is a well defined map z → ϕ(z) from
the cognitive tree to the space of I-states XI , body→mind field.
The mental process is described by iterations of body→mind field ϕ on
the (whole) cognitive tree. Such iterations are performed by, e.g., dynamical
system, (89).
Main cognitive features of the model:
a) Nonlocality of psychological functions.
b) Wholeness — integral evolution of the field ϕ.
c) Sensation — thinking. Since neural pathways go through the whole
body, a part of a pathway involved in a high level psychological function
can be connected to, e.g., skin-sensitivity. Thus high order psychological
functions also depend on various physiological stimuli.
d) Interrelation of distinct psychological functions. The central neuron S
of a cognitive tree plays the role of the center of the system of coordinates.
Other neurons can also be considered as such centers. Therefore the same
pathway contributes to distinct psychological functions.
e) Emotion based reasoning. Our pathway thinking model supports the
fundamental conjecture of Damasio, [52], that emotions play an important
role in the process of ‘reason–thinking’. Pathways going through centers
creating emotions can participate in a psychological function performing a
high order thinking process, e.g., proving of mathematical theorems. On the
other hand, pathways going through reasoning-centers can go through some
emotional center. Thus reason participates in the creation of emotions and
vice versa.

45 Dynamics in the Information Space

The body→mind field ϕ(z) describes important features of the functioning
of the neural system (in particular, its part located in the brain). However,
we will not be concentrated on the study of dynamics, e.g., (89), of the
body→mind field. The main reason is that (it seems) ϕ(z) describes merely
the production of information by the neural system (in particular, its part
located in brain) and not the flow of mental information by itself. I would
like to formulate this as:
Mental thesis: Mental activity is performed not in the pathway space,
the cognitive tree, but in the I-space.

143
 This thesis would be justified if there were obtained some neurophysiolog-
ical and cognitive science conformations. However, at the present time our
hypothesis on pathwaycoding of mental information looks quite speculative
from the neurophysiological viewpoint. In any case, at the moment there are
no experimental technologies that would give the possibility of measuring fir-
ings of neurons throughout even one long pathway of individual neurons. To
confirm our pathwaycoding hypothesis we have to measure simultaneously
firings of neurons for a huge ensemble of neural pathways.
Independence Thesis: The cognitive meaning (with respect to a psy-
chological function f ) of an I-state does not depend on a neural pathway that
produces this I-state.
Thus mental information does not remember its neurophysiological origin.
The Independence Thesis is supported (at least indirectly) by experimental
evidences that functions of some damaged parts of the brain can be (in some
cases) taken by other parts of the brain, see, e.g., [19], [52], [53]. This thesis
is also supported by neurophysiological evidences that very different neural
structures in brains of different species (e.g., fish and rat) can fulfill the
particular psychological function.
Of course, we should recall that by choosing the central neuron S we chose
the concrete psychological function f. Thus ‘the cognitive meaning’ is related
to this concrete psychological function. By choosing another psychological
function (a system of coordinates) we obtain another cognitive meaning.
In principle, the Independence Thesis might be considered as a kind of
anti-materialist thesis. We would not like to be at such a position. We un-
derstand well that the relation between the brain (in fact, in our pathway
model — the whole body) and mind plays the crucial role in mental activity.
The Independence Thesis should be considered as directed against the indi-
vidual deterministic relation between physical neural pathways in the body
and the cognitive meaning of the corresponding I-states. The absence of
such individual determinism does not contradict statistical determinism.
Thesis of Statistical Pathway Cognition. The cognitive meaning of
an I-state (with respect to a psychological function f ) is determined by the
statistical probability of realization of this I-state in the ensemble of pathways
Πf (the cognitive tree corresponding to f ).
Remark 2.1. (On the notion of probability) Over many years I studied
the foundations of probability theory, see [117]. I know well the great diver-
sity of viewpoints of the notion of probability. For me probability has nothing
to do with ‘potentiality’, ‘measure of belief’, and other perverse views. Prob-
ability is a statistical measure. Let E be a large ensemble of, e.g., physical

144
systems. Let these physical systems have some states. These states are rep-
resented as points in a state space. The probability of a state x with respect
to the ensemble E is given by the proportion:
the number of systems having the state x
p(x) = .
the total number of elements in the ensemble
In our model physical systems are centered neural pathways; states are strings
of firings throughout pathways — I-states. An ensemble E is a cognitive tree;
the state space is the I-space. We supposeThis is only a conjecture. that the
cognitive meaning of an I-state x ∈ XI is determined by the quantity
the number of neural pathways that produce x
p(x) = .
the total number of neural pathways in the cognitive tree
We call p(x) the mental state. Here and in all following considerations it is
assumed that a psychological function f is fixed. In fact, p(x) depends on
f : p(x) = pf (x).
Mental processes are probability–evolution processes. These are evolu-
tions of mental states. We have to find a mathematical model that would
provide the adequate description of the evolution: t → p(t, x).
It seems that mental processes should be described as discrete time evo-
lutions. We consider a discrete dynamical system in the space of probability
distributions:
p(tn+1 , x) = Lp(tn , x), (90)
where L is some operator in the space of probability distributions. The
generator of evolution L may be linear, may be nonlinear. There must be
performed experimental studies to find the form of L. Of course, L depends
essentially on a psychological function f and an individual.
Intuitively it is clear that there must be performed integration over the
whole I-space. The following evolution can be considered:
Z
p(tn+1 , x) = K(tn , x, y)p(tn , y)dy,
XI

where K(t, x, y) is a time-dependent kernel of evolution.

Continuous time models can be used as approximations of discrete-time
models.
The mental state p(t, x) is nothing other than the probability distribution
of the body→mind field ϕ(z). We can consider the cognitive tree Πf as a
probability space with the uniform probability measure
1
P(ω) =
number of elements in Πf

145
for ω ∈ Πf . Following to the probabilistic tradition we use the symbol ω
to denote points of the probability space. The map (body→mind field) ϕ :
Πf → XI is a random variable and the mental state

p(x) = P({ω ∈ Πf : ϕ(ω) = x})

gives the probability that neural pathways in the cognitive tree represent the
I-state x (or neural intensivity of the representation of x).
Thus the evolution of the mental state, t → p(t, x), can be reduced to
the evolution of the corresponding stochastic process ϕ(t, ω), the process of
body→mind correspondence.
But(!) probability theory tells us that we can not reconstruct the stochas-
tic process φ(t, ω) as a point wise map in an unique way on the basis of
corresponding probability distributions. The same mental flow can be gener-
ated by various neural flows. This trivial probabilistic argument gives strong
support to our Mental Thesis.
This probabilistic consideration can be used as a strong argument sup-
porting non-reductionism:
Neural reduction of mental processes is impossible.

46 Diffusion Model for Dynamics of Mental State

The simplest (but nontrivial!) model of mental evolution can be obtained by
considering Markovian body→mind fields. A Markovian process is a stochas-
tic process without long range statistical memory. In the Markovian case we
have the following equality for the conditional probabilities:

P(ϕ(tn+1 , ω) = xn+1 |ϕ(tn , ω) = xn , . . . , ϕ(t0 , ω) = x0 ) =

P(ϕ(tn+1 , ω) = xn+1 |ϕ(tn , ω) = xn ),

where t0 < t1 < . . . < tn < tn+1 . Here the I-state xn+1 at the instant of time
t = tn+1 is statistically determined by the I-state xn at the previous instant
of time t = tn . A Markovian mental state p(tn+1 , x) does not remember about
mental states at previous instances of time tn−1 , . . . , t0 .
Such a type of mental processing is natural for ‘low level’ mental activity,
e.g., reactions to stimuli. When we react at the instant of time t = tn+1 to
the state of hunger which we had at the instant of time t = tn we do not
recall all our states of hunger for the last few days or years. We perform just
a transition p(tn , x) → p(tn+1 , x). Here the mental state p(tn , x) is the state
of hunger and the mental state p(tn+1 , x) is the state of nourishingOf course,

146
previous experiences of hunger were used to build the cognitive tree corre-
sponding to nourishing. A part of this tree architecture is even transmitted
genetically..
We now consider a simpler reaction: to pain induced by fire at the instance
of time t = tn+1 . The mental state p(tn+1 , x), pain, does not depend even
on the previous mental state p(tn , x). Here the body→mind field produces
a sequence of independent random variables (e.g., the well known Bernoulli
process):

P(ϕ(tn+1 , ω) = x|ϕ(tn , ω) = y) = P(ϕ(tn+1 , ω) = x) ≡ p(tn+1 , x).

We now consider continuous time evolution for general Markovian

body→mind fields. The evolution of the mental state p(t, x) is described by
the Chapman-Kolmogorov equation:
∂p
(t, x) = Lp(t, x), (91)
∂t
lim p(t, x) = p0 (x),
t↓0

where L is the generator of the Markovian evolution. If we know the initial

mental state p0 (x) and the generator of mental evolution L we can find the
mental state at any instant of time t ≥ 0.
Remark 3.1. (Determinism or free will?) Our model of the process
of thinking gives the nice combination of mental determinism and free will.
Mental determinism is a consequence of deterministic evolution equations.
However, this determinism is a determinism of probabilities. Thus proba-
bilistic representation of the mental state is the origin of the feeling of free
will. We remark that the situation has some similarities with quantum the-
ory. Equation (91) is also called the direct Kolmogorov equation. Besides
this equation we can consider the inverse Kolmogorov equation:
∂p
(t, x) = L∗ p(t, x), (92)
∂t
lim p(t, x) = pT (x),
t↑T
∗
where L is the adjoint operator of the generator L of the Markovian evolu-
tion. If we know the mental state at some instant of time T > 0 we can recall
the mental state p(t, x) at any instant of time 0 ≤ t < T. Thus the inverse
Kolmogorov equation describes the process of recalling. It seems that human
beings have the ability to solve the direct and inverse Kolmogorov equations
(at least for small time intervals). This gives the possibility of predicting an

147
expected mental state by using the direct Kolmogorov equation and to re-
call a mental state from the past by using the inverse Kolmogorov equation.
Such an ability to solve the evolution equation for the mental state depends
essentially on a human individual.
We would like to remark that there exist stationary mental states p(x)
that are not changed in the process of evolution.
In the p-adic model of the I-space XI = Qp the simplest Markovian
evolution is given by the diffusion process (Vladimirov–Volovich diffusion).
This process was intensively studied in p-adic theoretical physics, see, e.g.,
[187]. The corresponding evolution equation has the form:
∂p 1
(t, x) = − Dx2 p(t, x), p(0, x) = p0 (x), (93)
∂t 2
where Dx is a kind of differential operator on the p-adic tree (Vladimirov’s
operator, [187], see chapter 5).
We can easily find a fundamental solution K(t, x). This is the solution for
the initial mental state p0 (x) = δ(x). Here δ(x) is the well known Diracs δ-
function. This is the probability distribution that is concentrated in the fixed
point. By knowing the dynamics K(t, x), x ∈ XI , of the mental state starting
with the ‘sharp mental state’ δ(x) (all neural pathways of the whole mental
tree produce the same I-state)We understood that such a mental state might
not be approached in reality. So K(t, x) is merely the ideal object used in the
mathematical model., we can find the dynamics of the mental state p(t, x)
for any initial distribution p0 (x) :
Z
p(t, x) = K(t, x − y)p0 (y)dy.
XI

47 Information Phase Space

We turn back to Vladimirov–Volovich diffusion, (93). The simplest way to
find the fundamental solution of this equation is to use the Fourier trans-
form on the p-adic I-space, see chapter 5 on the p-adic Fourier transform.
This transform determines a new mental representation which can be con-
sidered as dual to the mental position representation. Here we consider the
I-state variable x as the mental position variable. The dual variable ξ corre-
sponding to the new mental representation can be interpreted as the moti-
vation.motivation This is the motivation for changing the mental position(I-
state) that is described by the x-variable. This dual representation is called
a motivation representation,motivation representation see chapters 5 and 6.
We denote the dual mental space (space of motivations) by the symbol XI0 .

148
 By using the motivation representation for the fundamental solution K(t, x)
we can represent the dynamics of the mental state as
Z
p(t, x) = G(x, ξ)q(t, ξ)dξ,
XE

where q(t, ξ)dξ gives the dynamics of the probability distribution of motiva-
tions corresponding to the initial motivation state q(0, ξ) ≡ 1 on XI0 .
The probability distribution q(t, ξ) gives the statistical probability of neu-
ral pathways producing the motivation ξ :
the number of pathways producing the motivation ξ
q(t, ξ) = , (94)
the total number of pathways
where all quantities are related the fixed instant of time t.We remark that
q(t, ξ) = e−tH(ξ) , where H(ξ) is the free mental energy.
Remark 4.1 (On mental position and motivation representations) The
p-adic Fourier transform is one to one correspondence. Therefore from the
mathematical viewpoint the mental position state p(t, x) and the mental
motivation state q(t, ξ) provide equivalent descriptions of mental processes.
However, these representations are not equivalent from the cognitive view-
point. The I-space XI has a direct relation to the activity of neural pathways.
The mental state p(t, x) is the distribution of neural pathways producing con-
crete I-states x ∈ XI . We could not use the same picture for the motivation
state q(t, ξ). We cannot find the motivation ξ for the concrete neural path-
way. In some sense ξ is produced by the whole cognitive tree. We specially
wrote in (94) ‘pathways producing’ and not ‘pathways having’. One fixed
neural pathway can participate in producing of a few motivations. So the
‘function’ neural pathway → motivation is a multivalued map.
In chapter 6 there were presented motivations that the mental process
should to be described by using a mental phase spacemental phase space —
the product of the I-space and the motivation space: XI = XI × XI0 = Qp ×
Qp . In such a model we have to consider dynamics of probability distributions
p(t, x, ξ) on the phase space. Such dynamics can be described by the Liouville
equation on the p-adic phase space. By the analogy with ordinary statistical
mechanics we consider mental observables represented by functions a(x, ξ),
(with real values) on the mental phase space.
Results of measurements are given by mean values:
Z
hai = a(x, ξ)p(t, x, ξ)dxdξ. (95)
XI

These mean values are the only possible quantities that can be measured
over a cognitive tree. It seems to be impossible to prepare a cognitive system

149
in such a mental state that all neural pathways produce the same I-state x
and the same motivation ξ, see chapter 6 on the mental position–motivation
uncertainty relation. Thus variables x and ξ are kinds of hidden variables
and we could only measure averages over these variables.

48 Mental State as the Distribution of a p-adic

Random Walk
We have a collection Πf of centered neural pathways working for a psycho-
logical function f. Each pathway z belonging to (cognitive tree) Πf produces
an I-state x belonging to the I-space M. The mental state was defined as a
probability density of neural pathways producing a fixed mental point. One
of the simplest models of mental evolution is diffusion on the p-adic I-space
M . It is well known that ultrametric diffusion can be represented as a ran-
dom walk (S. Albeverio, W. Karwowski, X. Zhao, L. Brekke, M. Olson, S.
Evans, A. Figa-Talamanca,...), see, e.g., [187]. Such a representation was
widely used in models for spin glasses and proteins. It is natural to use a
random walk representation for the mental evolution, since in our model the
I-space has the p-adic ultrametric structure. First, we recall the ultramet-
ric (p-adic) random walk model for physical systems, e.g., spin glasses and
proteins, and then we propose an analogous mental model.
Consider a random walk with a p-adic space of states. Barriers separating
states are determined by the p-adic distance between states. States x and y
are separated by a barrier:
b(x, y) = b(|x − y|p ),
where b(s) is some function of the real variable. In general there is only one
restriction on the form of energy barriers:
A function b(s) monotonically increases when s = |x − y|p increases.
Thus points x and y which are located in a small ball are separated by a
small energy barrier. Therefore it is easy to move from, e.g., x to y. Points
x and y which are located far away from each other are separated by a high
energy barrier. Therefore the system jumps from, e.g., x to y not so often.
Such a behaviour is well described by transition probabilities of the form:
q(x, y) = q(|x − y|p ) = e−b(|x−y|p ) .
This model has been widely investigated in physical literature (for example,
by A. Ogielski and D. Stein; B. Huberman and M. Kerszberg; S. Gross-
man, F. Wegner and K. Hoffmann; R. Rammal, J. Angles d’Auriac and B.

150
Doucot; A. Blumen, J. Klafter and G. Zumofen). We shall use corresponding
mathematical results in our further mental investigations.
In the random walk mental model, instead of saying that at some moment
a neural pathway z produced an I-state x and at the next moment the neural
pathway z produced a new I-state y (as we did in previous sections), we say
that at the first moment the neural pathway z was located at the I-state x and
then z jumped to a new I-state y. In this model neural pathways randomly
walk over the I-space XI . The p-adic distance between mental points x and
y determines the mental energy barrier b(x, y) = b(|x − y|p ). To move from
an I-state x to an I-state y the neural path z must pass the mental energy
barrier b(x, y). As in the above physical model we introduce the probability
p(t, x) of the neural pathway being found in the mental point x. In previously
discussed model p(t, x) is nothing other than the mental state. The evolution
of p(t, x) can be described by the equation of ultrametric diffusion.
In physical literature there was intensively studied the behaviour of the
autocorrelation functions of states. We can use these results in our mental
model. Suppose that at time t = 0 a neural pathway z (working for a psycho-
logical function f ) is prepared in an I-state x (so the probability distribution
is concentrated precisely in x). We calculate the probability p0 (t) that after
a time t it will again be found in the same point x. The main distinguishing
feature of ultrametric diffusion is that (depending on the structure of mental
energy barriers) the autocorrelation function can have a power law decay:

p0 (t) ≈ t−γ , t → ∞.

For such a slow decay energy barriers bj must increase sufficiently quickly,
e.g., bj = jb, b > 0. Here γ = T ln p/b. In physical models the parameter T has
the meaning of temperature. In our model we call T the mental temperature.
If the mental temperature T is high then p0 (t) decreases very quickly. In this
case a mental pathway can not produce the same I-state for a long time.
The possibility of having a power law decay plays the important role in
the process of information exchange in the neural system. Suppose that p0 (t)
decreases very quickly — according to the famous Kohlrausch law:
β
p0 (t) ≈ e−t ,

with β = t/b. In such a case the mental state is very unstable. Its existence
time is not sufficient for information exchange between various psychological
functions. We recall that Kohlrausch’s decay can also take place for the
ultrametric random walk (if energy barriers do not grow fast enough, e.g.,
bj = b ln j).

151
 We recall that for finite n (where n is the length of a pathway) there is
the exact formula for the autocorrelation function p0 (t). In principle, this
formula can be tested experimentally — for a fixed neural pathway of the
length n. If this formula is confirmed experimentally, it could be used to find
the mental temperature T (T = Tf depending on a psychological function
f ). The mental temperature is an important macroscopic characteristic of
a psychological function. The mental temperature T can be also connected
with transition probabilities. The probability of jumping over the mental
energy barrier bm is equal to qm = e−bm /T .
We remain that in the model of the ultrametric random walk studied in
the physical literature the transition probability increases with decreasing of
the p-adic distance. Such a behaviour has interesting mental consequences.
First we consider the output part of a neural pathway z. It is essentially
easier to change the states of neurons which are located (on the z) farther
from the central neuron S than neurons which are located (on the z) near
the S. If we change only states of far neurons then the distance between cor-
responding mental points is still small. Thus such a transition can be easily
performed, since the mental energy barrier b(x, y) = b(|x − y|p ) decreases,
when |x − y|p → 0.
It is clear that for the input part of the neural pathway z the behaviour is
completely different. Here the S has very little influence on neurons (located
on the input part of the pathway z) which are far from it.
I think that such a behaviour does not contradict neurophysiological and
cognitive data.
The most important consequence of our mental model is that a psycho-
logical function f has the largest influence on domains of the brain which
are spatially separated from the central neuron S = Sf of the f. This might
explain the greatest mystery of the brains functioning — spatial separation
of domains of concentration of a few psychological functions performing some
synchronized task; see, for example, [19], p. 65:
“Visual information flows from the eyes to the thalamus, then to V1 and up,
ultimately to perhaps forty separate areas for visual analysis. The first five visual
areas, V1 to V5, are believed to be deeply involved in visual consciousness.”
Thus the ultrametric p-adic topology on the I-space determines this spa-
tial separation of domains of concentration of psychological functions. From
the physical point of view this spatial separation is, on the one hand, a con-
sequence of the neural pathway coding of mental information and, on the
other hand, the random walk structure of the functioning of psychological
functions.

152
49 Discussion of the Neural Pathways Think-
ing Model
6.1. Can consciousness be treated as a variable?
The first part of the book [19] of B. J. Baars contains an interesting
discussion on the possibility of treating consciousness as a variable and the
importance of such a treatment in cognitive science. My attempt to quantify
consciousness was, in particular, motivated by this discussion. However, we
essentially modified Baars’ idea about consciousness variable. As I under-
stood it, his consciousness variable would be a kind of Newtonian physical
variable such as position, velocity, force; see [19], p. 11, on comparison of the
consciousness variable with Newtonian gravity. Our consciousness variable,
mental state, is a probability distribution. Thus there are more similarities
with statistical mechanics and even with quantum mechanics. Such a variable
is not a local variable on the brain. This is merely a wholeness variable.
Starting with a mental state we can define a quantative measure of con-
sciousness, a kind of ‘temperature of consciousness’.temperature of conscious-
ness First, we discuss the connection between levels of neural activity and
levels of consciousness. The idea that there is a direct correspondence be-
tween levels of neural activity and levels of consciousness is a very common
postulate in cognitive science. An extended discussion of this problem can
be found, for example, in [19], pp. 18–19. I disagree that there is such a
direct relation between levels of neural activity and levels of consciousness.
For example, let us consider the extreme case in which all possible states
of the brain are activated. Such a super-activation definitely would not imply
a high level of consciousness. It might be that it is not simply the level of
neural activity that determines the level of consciousness.
Our conjecture is that consciousness is determined by the variation of a
mental state (compare to chapter 6). One of possible numerical measures of
the level of consciousness is the entropyentropy of a mental state. It is well
known that (for the discrete probability distribution) entropy approaches the
maximal value for the uniform probability distribution. In our interpretation
this is the lowest level of consciousness (so it may be better to use entropy
with a minus sign as a quntative measure of consciousness).
Another possibility is to define a measure of consciousness as the variation
of a mental state (with respect to the p-adic metric on the I-space). It seems
to be a more natural measure of consciousness, since the topological structure
of the I-space must be taken into account. The p-adic variation can be defined

153
by using Vladimirov’s differential operator [187] D :
Z
C(p) = Consciousness(p) = |Dp(x)|2 dx, (96)
XI

where p(x) is the mental state of a cognitive system. Consciousness is al-

ways non-negative and it takes its minimal value, C(p) = 0, for the uniform
probability distribution. In our interpretation this is the lowest level of con-
sciousness. At the moment we do not know anything about mental states
producing consciousness at the extremely high level, namely, solutions of the
problem: C(p) → max.
6.2. Are animals conscious?
Our model strongly supports the hypothesis that animals are conscious
(see, e.g., [19] on the detailed neurophysiological and behavioral analysis of
this problem). If consciousness is really determined by the variation of the
mental state then animals are definitely able to produce such nontrivial vari-
ations by their systems of neural pathways. Moreover, in our Neural Pathway
Model pathways going throughout the body play an important role in the
creation of consciousness. Thus the role of differences in brain structures
should not be overestimated, compare with [19]. On the other hand, animals
have lower levels of consciousness, C(p). We can quantify this problem in the
following way. There exists a threshold Chuman . Animals could not produce
mental states p(x) such that C(p) is larger than the human consciousness
threshold Chuman . And human beings (at least most of them) could produce
(at least sometimes) mental states for that consciousness is larger than this
threshold.
6.3. Blindsight.
Our model might be used to explain the mystery of blindsight and similar
phenomena [19]:
“The mystery of blindsight is not so much that unconscious visual knowledge
remains.... The greatest puzzle seems to be that information that is not even
represented in area V 1 is lost to consciousness when V 1 is damaged.”
However, in our Neural Pathway Model by destruction of some individual
neurons (e.g., in the area V 1) we destroy (modify) huge ensembles of neural
pathways. Of course, in our model information was never preserved in the
area V 1 nor in some other localized area. Information is preserved by en-
sembles of pathways, and they are not located in some particular domain of
brain.
However, we also have to explain the unique function of area V 1 in the
creation of visual consciousness [19]:

154
 “V 1 is the only region whose loss abolishes our ability to consciously see objects,
events... But cells in V 1 respond only to a sort of pointillist level of visual percep-
tion... Thus it seems that area V 1 is needed for such higher-level experiences, even
though it does not contain higher-level elements! It seems like a paradox.”
I agree that the area V 1 is the unique region that damage destroy our
ability to conscious visualization. But I recall that our model is, in fact,
the Centered Neural Pathway Model. The uniqueness of area V 1 in con-
scious visualization is determined by the central neurons of cognitive trees
corresponding to the psychological function of conscious visualization being
located in area V 1.
We continue the quotation from Baars [19]:
“Cells that recognize objects, shapes, and textures appear only in much ‘higher’
regions of cortex, strung in a series of specialized regions along the bottom of the
temporal lobe.”
Yes, these cells are centers of cognitive trees corresponding to other psy-
chological functions, object recognition.
6.4. Non-Markovian evolutions of the mental state.
We have described in detail Markovian models of mental evolution. Such
evolutions are quite natural for low-level mental processes, in that long range
memory is not involved. However, higher-level mental processes are defi-
nitely non-Markovian. In the opposite of the Markovian case, by knowing
the mental state (probability distribution) at the previous step and transition
probabilities, in general we can not determine the next mental state. Unfor-
tunately non-Markovian evolution is essentially more complicated from the
mathematical viewpoint than Markovian evolution.
6.5. Role of loops.
It is well known that neural loops play an important (but still rather
unclear) role in the ‘production’ of consciousness (e.g., in Edelman’s theory,
see also [19] on the extended discussion). We consider this problem from
the purely informational viewpoint by using the Neural Pathway Model. We
shall see that the presence of loops in a cognitive tree incredibly increases
the informational power of a tree.
Let us consider a tree without loops. Now let us add to this tree a single
loop going through the central neuron S. This implies not just adding one
new neural pathway (which can produce only one I-state). This single loop
should be added (with a few cycles) to every existing pathway. Thus every
previously existing pathway is preserved in a new tree, but each such pathway
also produces a huge number of new pathways corresponding to additional
cycling around the additional loop.

155
 Thus by adding just a single new loop we change essentially a mental state
– the probability distribution of firings throughout pathways. Now imagine
that a few new loops are added to the old tree! By using loop-pathways the
brain can increase the informational power in a very economical way.
6.6. Neural code and the structure of I-space.
Suppose that the coding system of a cognitive system is based on a fre-
quency code. There exists a time interval ∆ (depending on a cognitive system
and a psychological function) such that an I-state produced by a centered
neural pathway is a sequence with coordinates given by numbers of oscilla-
tions for corresponding neurons during the interval ∆. Thus in our model the
problem of the neural code is closely related to the problem of time-scaling
in the neural system. For different ∆ we obtain different coding systems,
and, consequently, different structures of I-spaces. The corresponding natu-
ral number p that determines the p-adic structure on the I-space is defined as
the maximal number of oscillations that could be performed by neurons (in
the cognitive tree for some fixed psychological function) for the time interval
∆. The frequency coding based on the 2-adic system induces the 2-adic I-
space that differs crucially from the 5-adic (or 1997-adic) I-space induced by
the 5-adic (or 1997-adic) system. As was remarked in chapters 1 and 3, by
changing the p-adic structure we change crucially dynamics. Hence the right
choice of the time scaling parameter ∆ plays the important role in the cre-
ation of an adequate mathematical model for functioning of a psychological
function.
6.7. Psychological time.
There might be some connection between the time scale parameter ∆
of neural coding and psychological time.psychological time There are strong
experimental evidences that a moment in psychological time correlates with
≈ 100 ms of physical time for neural activity. In such a model the basic
assumption is that the physical time required for the transmission of infor-
mation over synapses is somehow neglected in psychological time. In the
model, the time (≈ 100 ms) required for the transmission of information
from the retina to the inferiotemporal cortex (IT) through the primary vi-
sual cortex (V1) is mapped to a moment of psychological time. It might be
that by using ∆ ≈ 100 ms we shall obtain the right p-adic structure of the
I-space.
Unfortunately it seems that the situation is essentially more complicated.
There are experimental evidences that the temporal structure of neural func-
tioning is not homogeneous. The time required for completion of color infor-
mation in V4 (≈ 60 ms) is shorter than the time for the completion of shape
analysis in IT (≈ 100 ms). In particular, it is predicted that there will be

156
under certain conditions a rivalry between color and form perception. This
rivalry in time is one of the manifestations of the complex level temporal
structure of brain. It can be shown that at any given moment in physical
time there are neural activities in various brain regions that correspond to a
range of moments in psychological time. In turn, a moment in psychological
time is subserved by neural activities in different brain regions at different
physical times.
Therefore it is natural to suppose that different psychological functions
have different time scales, and consequently different I-spaces. Thus one
psychological function is based on the 2-adic I-space and another on the
5-adic (or 1999-adic) I-space. This is the very delicate point and we shall
clarify it.
There is the total space Π of all neural pathways. The concrete psycho-
logical function f is based on some centered cognitive tree Πf (a subset of Π).
There exists the fixed time scale ∆ = ∆f corresponding to this psychologi-
cal function. Hence there exists a natural number p depending on ∆f (and
hence on f ) determining the p-adic structure of the mental space for f. Thus
p = pf . On this pf -adic I-space there is defined the mental state pf (x) of the
f. In general another psychological function g has its own time scale ∆g and
corresponding pg . Its mental state pg (x) is well defined on the pg -adic space.
If pf is not equal to pg (e.g., pf = 2 and pg = 1997) then dynamics of mental
states corresponding to psychological functions f and g differs crucially —
even if evolutions are described by the same (e.g., diffusion) equation.
Finally, we remark that many psychological functions are strongly interre-
lated on the neural pathway level. Cognitive trees Πf and Πg corresponding
to psychological functions f and g can have a large intersection. In the
extreme case these trees could even coincide: Πf = Πg . But the use of dif-
ferent time scales ∆f 6= ∆g would produce totally different evolutions for
corresponding mental states.
6.8. Discreteness of time.
Previous considerations demonstrated that the continuous real time model
for the evolution of the mental state gives only rough approximation to the
really performed discrete time evolution. Moreover, discretization steps de-
pend on corresponding psychological functions.
Thus the evolution of the mental state pf (t, x) of a psychological function
f is described by discrete time dynamics, for example, (90), where tn+1 =
tn + ∆.
6.9. What is about our p-adic structures?
If we accept Edelman’s Neural Darwinism then we have to consider the
possibility that the p-adic structures of our psychological functions can de-
pend on individuals. Thus for an individual I the basis of his/her I-space

157
for a psychological function f depends both on f and I : p = pf,I . For two
different individuals, for example Ivan and Andrei, the same psychological
function f can be based on two different I-spaces:XIIvan = Qp and
f,Ivan

XIAndrei = Qp , respectively.
f,Andrei
Therefore the evolution of the same type (e.g., diffusion) would produce
completely different mental states (even if the initial mental state were the
same, e.g., the uniform probability distribution).
[Contextual Approach to Quantum Theory] Contextual Approach to
Quantum Theory
Often we enter the unknown edifice of a new scientific discipline through
a lesser gate that leads us into a side passage. It may take us a long while to
find our way to the main portal and to view the whole structure in its proper
perspective. E. Schrödinger.
It is all mysterious. And the more you look at it the more mysterious it
seems. R. Feynman.
In this chapter we discuss the recently developed contextual probabilistic
interpretation of quantum theory, see [123], [124], Växjö interpretation.Vaxjo
interpretation Oppositely to the Copenhagen interpretationCopenhagen in-
terpretation (at least its orthodox form) the Växjö interpretation is a real-
istrealist interpretation of quantum theory. The possibility of using such a
realist interpretation of quantum theory is especially important in applica-
tions of quantum formalism to cognitive sciences.
As has already been remarked in chapter 5, there are no reasons to believe
in the possibility of reducing mental processes to processes that occur in the
microworld, the ‘quantum world’.Unfortunately, the term ‘quantum’ is rigidly
reserved for micro objects. Even people speaking about quantum macro ef-
fects, e.g., Bose–Einstein condensate, have in mind collective behaviour of
elementary particles. Therefore we would use the term ‘quantum-like’ for
quantum models of macroscopic objects. Quantum-like behaviour need not
be directly related to collective behaviour of elementary particles. Person-
ally I do not believe that such things as the quantum dynamics of atoms,
electrons, photons, collapses of states with mass-superposition, wave–particle
dualism, the many world interpretation have any direct relation to mental
processes and, as it is claimed by many quantum physicists, to the phe-
nomenon of consciousness.
It is well confirmed by neurophysiological and cognitive investigations
that mental processes are performed at the macro level, namely, via neural
networks. Of course, all macro objects (including neural networks) are com-
posed of elementary particles. However, it would be too naive to believe that
the dynamics of elementary particles in the brain (or, more precisely, brain–

158
body, see chapter 8) neural networks might determine mental dynamics. We
should take into account the synergetic viewpoint.synergetic viewpoint Nowa-
days it is well known that a complex system can have new properties which
are not predetermined by properties of parts of the system. And on this occa-
sion it would be very useful to recall Charles Darwin’s words: “consciousness
seems to be the effect of complexity of organization.”
For example, in every table there are atoms, electrons, photons, ..., and
there occur collapses of mass-superpositions (of course, depending on the in-
terpretation of quantum theory). However, a table is not a cognitive system.
By some synergetic reasons all these quantum processes in a table could not
induce mental processes! Therefore when I speak about quantum-like mental
models I do not have in mind the possibility of quantum microscopic reduc-
tion. Everything in the mental game is macroscopic! As was pointed out in
chapter 5, by talking about quantum-like behaviour of mental systems we ex-
press the possibility that mental probabilistic behaviour might be described
by the quantum probabilistic formalism. The use of such a formalism has
nothing to do with wave–particle dualism, collapse, Planck’s constant, and
so on.
As was recently discovered in [123], [124] the quantum probabilistic for-
malism (Hilbert space formalism) can be reduced to the calculus of context
dependent probabilities. If we take into account dependence of probabilities
on context, then we immediately find that the classical calculus of probabili-
ties (based on Kolmogorov’s axiomatics for the measure-theoretical approach)
can be violated for some contexts. In particular, context dependence can pro-
duce interference of probabilitiesinterference of probabilities of alternatives.
Physical contextPhysical context is the complex of all experimental phys-
ical conditions under which a measurement is performed. Cognitive con-
textCognitive context includes the physical one: in particular, the structure
of neural networks in brain and body and physical, e.g., electric, processes
that take place in these neural networks. However, cognitive context is not
reduced to the physical one, see later considerations.
One of the main distinguishing features of the contextual probabilistic
interpretation, the Växjö interpretation, of quantum mechanics is the pos-
sibility of a realist description. In opposition to the Copenhagen interpre-
tation, in the Växjö interpretation there are no constraints that imply the
impossibility of realist deterministic models that would reproduce quantum
probabilistic behaviour. In the Copenhagen interpretation quantum random-
ness is considered as fundamental and irreducible randomness, see, e.g., [201].
It could not be reduced to randomness of initial conditions and mutual dis-
turbances, see the dialogue between the author and J. Summhammer [181]
for the details. We shall discuss this problem in more details. At the mo-

159
ment we just point out that such irreducible randomness looks quite counter
intuitive. We have no idea about possible probabilistic roots of such random-
ness. People say :‘this is quantum reality’... This irreducible randomness is
the root of all quantum mystifications. In particular, this irreducibility of
randomness is the main constraint (in the Copenhagen interpretation) which
implies the impossibility of constructing a realist deterministic model that
would reproduce quantum randomness. Roughly speaking Niels Bohr and
Werner Heisenberg (and a few generations of quantum physicists after them)
used the following considerations:
Look at the statistical data which produces interference. This data can
not be described by the ordinary ‘classical’ probability theory. But every
realist deterministic model (in which particles move along trajectories) will
produce statistical data described by ordinary probability theory. Therefore
such a realist basis of quantum theory can not be constructed.
Our notion of cognitive context is closely related to the notion of context
which is used in cognitive sciences. A cognitive context includes not only
a physical context, but all unconscious mental processes performed in the
brain and body. Thus cognitive context is physical context + unconscious
mental information, cf. [19], p. 145:
“Contexts are after all, largely unconscious intentions and expectations which
have been stable so long that they have faded into the background of our lives. We
take them for granted, just as we take for granted the stability of the ground on
which we walk.”
One of the quantum-like cognitive models is the model in which self is
simply a measurement device which performs measurements of various men-
tal observables over various cognitive contexts. Such a model has the main
distinguishing feature of the quantum model. Cognitive context is very sensi-
tive to acts of conscious measurements. When my self measures some mental
observable I, the image of some person (my lovely wife), such an act of mea-
surement perturbs essentially the cognitive context. Suppose that at the next
moment my self measures another mental observable J, the image of a car
which is making a dangerous maneuver in front of my car. The result of this
measurement of J can differ essentially from the result of the measurement
of J in the absence of the previous measurement of I. In the Hilbert space
formalism this sensibility to perturbations is mathematically expressed via
the noncommutativity of the corresponding mental observables. Of course,
it is well known that the observer metaphor for self was strongly criticized in
psychology and the cognitive sciences, see Ryle [161] and Baars [19], p. 143:
“In 1949, the Oxford philosopher Gilbert Ryle famously pointed out a contra-
diction in the common notion of ‘the self as observer’. He thought it made no

160
sense to postulate an observing self because it does not explain anything at all, it
merely moves the job of explanation to another level. If we had an observing self
contemplating the contents of consciousness, he argued, how would we explain the
self itself? By another observer inside the inner self? That would lead to a infinite
regress of observing selfs each looking into the mind of the preceding one, little
imaginary men inside our heads observing each other’s observations. The observ-
ing self — the homunculushomunculus, or, little human — seems to be a fallacy of
common sense. Ryle’s arguments against this ‘host in the machine’ has persuaded
countless scientists and philosophers that ‘the self’ is a snare and a delusion.”
We have no possibility of discussing this hard philosophical problem in
this book. We simply refer to Baars’ book [19], chapter 7, p. 145, which
contains strong arguments in favour of the ‘self as observer’:
“We will take the notion of an observing and agentive self as useful as long as
the job of explanation is not simply handed on to another level...”
By using the direct analogy with quantum measurements we understand
that it seems to be possible to escape Ryle’s infinite tower of observers and use
the notion of self-observer. Of course, we should use the Växjö interpretation
(the contextual statistical realist interpretation) of quantum mechanics. If
we try to use the Copenhagen interpretation and, consequently, associate a
measurement with the collapse of a wave function, then (even in the ordinary
theory of physical measurements) we shall obtain the same problem of the
infinite tower of observers, see Ballentine, [22], p. 369:
“Von Neumann’s theory of measurement is very unsatisfactory... But this can
only lead to an infinite regression, which, in an earlier era might have been taken
as a proof of the existence of God (the Ultimate Observer!)”.
p-adic quantum models developed in chapters 5 and 8 can be used for
the mathematical representation of ‘self’ (as observer). The same formalism
can be used for external measurements over cognitive systems which are
performed by, e.g., psychologists or neurophysiologists. On the other hand,
using the formalism of chapter 6 (p-adic-valued wave functions) for describing
external measurements is rather questionable. We recall that in the latter
formalism probabilities are not ordinary (Kolmogorov) real probabilities, but
p-adic ‘probabilities’. On the other hand, it might be that ‘self’-observations
can produce such unusual probabilities.
This chapter contains three sections which are devoted to fundamental
questions of quantum theory: interpretations and the possibility of creat-
ing local deterministic models reproducing quantum randomness, contextual
probabilistic calculus as the basis of quantum theory; a fundamental principle
for quantum theory (an analogue of Einstein’s principle).

161
 In the first section there are presented critical arguments against the indi-
vidual interpretation of Bohr’s complementarity and Heisenberg’s uncertainty
principles. The statistical interpretation of these principles is discussed in
the contextual framework. We support the possibility of using the statistical
contextual realist interpretation of the quantum formalism as well as other
physical formalisms. In the second section we study the role of context, a
complex of physical conditions, in quantum as well as classical experiments.
It is shown that by taking into account contextual dependence of experi-
mental probabilities we can derive the quantum rule for the addition of the
probabilities of alternatives. Thus we obtain quantum interference without
appealing to the wave (or Hilbert space) approach. The Hilbert space rep-
resentation of contextual probabilities is obtained as a consequence of the
elementary geometric result: cosine-theorem. By using another result from
elementary algebra we obtain the complex amplitude representation of prob-
abilities. Finally, we find the contextual origin of noncommutativity of in-
compatible observables. In the third section we find a fundamental principle,
the law of statistical balance in nature, which specifies quantum statistical
theory amongst all other statistical theories of measurements. This princi-
ple plays in quantum theory the role that is similar to the role of Einstein’s
relativity principle.
The contextual approach to probability, information, and quantum the-
ory which will be presented in this chapter has been intensively discussed
with many physicists and mathematicians. I would like to thank all them for
those fruitful (critical) discussions. I am especially grateful to S. Albeverio, L.
Accardi, L. Ballentine, V. Belavkin, E. Beltrametti, G. Cassinelli, A. Cheb-
otarev, W. De Baere, W. De Myunck, R. Gill, D. Greenberger, S. Goldstein,
C. Fuchs, L. Hardy, A. Holevo, T. Hida, B. Hiley, P. Lahti, E. Loubents, D.
Mermin, T. Maudlin, A. Peres, I. Pitowsky, A. Plotnitsky, A. Shiryaev, O.
Smoljanov, J. Summhammer, L. Vaidman, I. Volovich, A. Zeilinger.

50 The Växjö Interpretation of Quantum Me-

chanics
We start with critical arguments against the individual understanding of
Bohr’s complementarity and Heisenberg’s uncertainty principles. It is ar-
gued that the only possible interpretation of these principles is the statistical
contextual interpretation. In particular, oppositely to orthodox Copenhagen
views such an interpretation does not imply the impossibility of assigning
to elementary particlesThe reader may protest against the use of the notion

162
‘particle’. It might be better to use ‘quantum system’. But I dislike using
‘quantum’ regarding real physical objects. ‘Quantum’ has become merely
symbolic of using a rather special mathematical formalism. objective proper-
ties; including position and momentum. Of course, such an ontic description
of physical reality might not be provided by the quantum formalism. In par-
ticular, it is clear that the quantum formalism does not provide a complete
description of physical reality.
We argue that the most natural interpretation of physical theories (in
particular, the quantum formalism) is the statistical contextual realist inter-
pretation — the Växjö interpretation.
1.1. The contextual statistical realist interpretation of physical
theories.
Quantum theory, as well as classical physics, is used to describe properties
of the pairs
π = (a physical system, measuring device). (97)
I do not think that the understanding of this notion, the contextual structure
of physical theories, was really Bohr’s invention. It was clear to everybody
that physical observables are related to properties of a physical systems as
well as of measuring devices. The main invention of Bohr was not contextu-
ality, but complementarity. Bohr’s greatest contribution was the recognition
that there exist complementary experimental arrangements and hence com-
plementary, incompatible, pairs π1 , π2 of the form (97). I think nobody can
be against the recognition of such a possibility. Why not? Why must all
contexts, complexes of physical conditions, be coexisting? Contextuality and
complementarity are two well understandable principles (not only of quan-
tum physics, but physics in general).
The real problem was that Bohr as well as Heisenberg (and further gener-
ations of their adherents) did not pay attention to quantum complementarity
being the experimental fact concerning pairs

π = (elementary particle, macroscopic measuring device) (98)

and not elementary particles by their selves. It is a pity that the greatest
promoters of contextualism forgot about the contextual basis of complemen-
tarity.
If we remember about contextuality in discussing complementarity we
see that complementarity of contexts in quantum physics does not at all
imply complementarity of corresponding objective properties (of elementary
particles) contributing to such observables. In particular, contextual com-
plementarity does not at all imply that elementary particles do not have

163
objective properties. In particular, there are no reasons to suppose that it
is impossible to provide a kind of hidden variable (HV) description, an ontic
description, for these objective properties. Mathematically the pair π can be
described by two variables λs and λd representing objective states of a phys-
ical system and a measuring device. So a physical observable is represented
as a function
A = A(λs ; λd ). (99)
This is the mathematical representation of realism.
In general, λs can be represented as a vector with numerical coordinates.
Numerical does not mean ‘real number’. It can be as well, e.g., a p-adic
number, see [187], [109], [111]. So the possibility of the HV-description does
not imply continuous infinitely divisible representation of physical reality.
In principle, the classical physical description is often identified with the
description based on the real number ‘continuous’ model. We would not
like to use such a restrictive, XVIIIth, XIXth century, and first half of the
XXth century interpretation of classical physics. A classical physical model
is every model that provides an objective description of phenomena – the
representation of physical observables in form (99). In particular, a classical
description need not be identified with the description given by Newton’s dif-
ferential equations. I think that the mixing of ‘classical’ and ‘continuous’ was
one of main roots of the vague interpretation of results of experiments with
elementary particles. Thus the development of alternative (non-real, noncon-
tinuous) classical models, e.g., p-adic or discrete, might play an important
role in the clarification of the foundations of quantum theory.
On the other hand, an adherent of Bohr would argue (see A. Plotnitsky
in [121]) that
“Such a separation, and hence the description of (properties of) quantum ob-
jects and processes themselves (as opposed to certain effects of their interaction
with measuring instruments upon the former) are impossible in Bohr’s interpreta-
tion.”
I think that the origin of such an interpretation of complementarity by
Bohr was the individual interpretation of Heisenberg’s uncertainty principle:
h
∆q∆p ≥ . (100)
2
The close relation between Bohr’s complementarity principle and Heisen-
berg’s uncertainty principle is well known. For two years (1926–1927) Bohr
could not present any model that could explain both corpuscular behaviour
(black body radiation, photoelectric effect) and wavelike behaviour (the two
slit experiment, diffraction) of elementary particles. It is well known how

164
heavy this thinking process was for Bohr — really a kind of mental disaster.
Only after the derivation by Heisenberg of the uncertainty principle, did Bohr
propose a new model of physical reality based on the principle of complemen-
tarity. Unfortunately, Heisenberg’s uncertainty relation was interpreted as
a relation for an individual elementary particle. The main problem was the
mixing by Heisenberg of individual and statistical uncertainty. For example,
in his famous book [89] he discussed the uncertainty principle as a relation for
an individual system, but derived this principle by using statistical methods!
The roots of such individual complementarity can already be found in
the first article of Heisenberg. At the beginning Heisenberg’s quantization
procedure was not statistical. It seems that Heisenberg was sure that he
had found equations for observables related to individual physical systems.
It was a quite common point view: in classical mechanics the position and
momentum of an individual (!) system are described by real numbers, and
in quantum mechanics by matrices.
I think that the starting point of such an individual viewpoint was a per-
verse understanding of probability by Heisenberg. He identified probability
with potentiality. And, of course, potentiality might be assigned to an indi-
vidual particle. However, ‘potentiality–probability’ could not be well defined
in a rigorous framework of natural science. Probability should be assigned
not to an individual object but to an ensemble of objects. So probability
is not potential, but statistical. In fact, Heisenberg’s views were close to
Bayesian views to probability as the measure of belief. It is interesting to
remark that modern adherents of Heisenberg’s viewpoint, e.g., C. Fuchs, (the
so called information interpretation of quantum mechanics, see [121]) imme-
diately recognized that the only probabilistic possibility of proceeding in this
direction is to use a Bayesian approach.
Heisenberg rightly underlined that the matrix-description could not be
used for describing the objective position and momentum of an electron.
There (as everywhere in XVIIIth–XXth centuries physics) objective reality
was, in fact, identified with the continuous (real number) mathematical model
of physical reality. The impossibility of creating a continuous real number
model for the motion of an electron in Bohr’s model of the atom was consid-
ered by Heisenberg (and many others) to mean it was impossible to assign
objective properties to electrons. It was a rather strange passage. But we
understand that it was the beginning of the XXth century and Heisenberg
used the standard mathematical image of physical reality — the continuous
real number model.
I think that the absence of a continuous classical model for the motion
of an electron in Bohr’s atom does not imply the impossibility of creating
other, noncontinuous, classical (causal deterministic) models. Moreover, the

165
original considerations of Heisenberg did not even imply the impossibility of
creating a continuous classical model — as was claimed by Heisenberg and
then by Bohr. The story is much simpler: first Bohr tried to create such a
model, but could not. Then Heisenberg — with the same result. After this
it was claimed that such a model did not exist at all. And what is most
interesting — not only for Bohr’s model of the atom (well might it be), but
for any other model... I cannot understand this kind of ‘quantum logic’.
Actually Heisenberg proposed a mathematical model for some class of
observations of position and momentum of electron. These observations sat-
isfy the uncertainty relation. It is not clear why we cannot present other
mathematical models for some other observations of position and momen-
tum of an electron which would violate this relation. Of course, if we relate
Heisenberg’s position and momentum to an individual electron then such
individualization plays the role of the objectification of these quantities. It
is rather strange logical circle, but it seems that it was done by Heisenberg
and Bohr. Finally, this objectification in combination with the uncertainty
relation implies (for Heisenberg and Bohr) the impossibility to consider other
position and momentum variables, distinct from Heisenberg’s variables. On
the other hand, if we use the statistical interpretation of the uncertainty re-
lation then there are no reasons for such ‘no-go’ conclusions. Well, we could
not prepare statistical ensembles with small dispersions for two variables in-
troduced by Heisenberg. But we should have great imagination to make the
Heisenberg-Bohr conclusions.
Another important story that stimulated Bohr’s complementarity think-
ing was Schrödinger’s quantum story. It is important to recall that Schrödinger
was sure that he had discovered a completely new method of quantization,
see [112]. It is well known that many famous physicists had great prejudice
against Heisenberg’s approach to quantization. So Schrödinger’s wave me-
chanics was considered by many of them to be the end of quantum mechanics;
as the possibility of describing ‘quantum experiments’ by using the classical
theory of partial differential equations (especially strong comments against
Heisenberg–Bohr were made by Einstein and Wien).
Hence Heisenberg and Bohr had to find some strong arguments in favour
of Heisenberg’s approach or disappear from the quantum scene. Moreover,
the whole quantum spectacle could have ended with quite trivial final: in-
stead of a great mystery, we could simply obtain a part of the well established
theory of partial differential equations. If all those circumstances were to be
taken into account it would be clear how Bohr created the complementarity
principle with all its ‘no-go’ consequences.
Hence Bohr’s complementarity was a kind of individual complementarity.
Complementary features were related to an individual physical system. It is

166
a pity that the contextualists Bohr and Heisenberg related the uncertainty
relation not to some special class of measurement procedures of the position
and momentum described by the quantum formalism, but to the position
and momentum of an individual elementary particle. This implies the pre-
judgments that the position and momentum even in principle could not be
determined simultaneously. Moreover, it is even in principle impossible to
assign such a physical property, e.g., position or momentum, to, e.g., an
electron: an electron does not have a trajectory.
In fact, the only possible interpretation of Heisenberg’s uncertainty prin-
ciple is the statistical contextual interpretation, see, e.g., [21], [22]. It is
impossible to prepare such an ensemble of elementary particles in which dis-
persions of both position and momentum observables would be arbitrary
small. Everybody would agree that this statement can only be verified ex-
perimentally. Contextualism has to be statistical contextualism, and con-
sequently complementarity has to be statistical contextual complementarity.
Such contextualism and complementarity do not contradict the possibility of
finer description of reality than given by quantum theory.
The complex of experimental physical conditions must be split into two
complexes — a preparation procedure and a measurement procedure, see,
e.g., [21]. A preparation procedure produces a statistical ensemble of phys-
ical systems. Then a measurement device produces results — values of a
physical observable. Nonzero dispersion of this random variable does not
imply that the individual physical system does not have objective properties
which produce values of the observable via (99). Thus contextual statisti-
cal interpretation can be, in principle, extended to the contextual statistical
realist interpretation.
Of course, individual complementarity can be used as an argument against
the possibility of creating a finer description of physical reality than that
given by quantum mechanics. But statistical complementarity cannot be
used as such an argument. By contextual statistical interpretation it is not
forbidden in principle to create preparation and measurement procedures
such that position and momentum would be measured with dispersions ∆q
and ∆p such that ∆q∆p < h/2. Of course, such a statement should imme-
diately induce a storm of protests about the principle of complementarity.
However, we again recall that the correct complementarity principle is con-
textual and statistical. It is about some class of measurement and prepara-
tion contexts described by the quantum formalism. In particular, we do not
consider the quantum formalism as a kind of complete physical theory.
However, in so far as we cannot perform such experiments for elementary
particles, it is actually impossible to reject Bohr’s principle of individual
complementarity Well, we may produce various realist models underlying

167
the quantum formalism, for example, Bohmian mechanics, stochastic elec-
trodynamics. However, it seems that all such models would be more or less
automatically rejected by the quantum community.. What can we do?
We can study consequences of the general statistical contextuality and try
to demonstrate that some distinguishing features of quantum theory (which
are typically associated with individual complementarity, ‘no-go’ complemen-
tarity) are, in fact, simple consequences of statistical complementarity, ‘go
deeper’ complementarity. I did this in a series of papers, see, e.g., [123], [124]
and also section 2 of this chapter for the final version. The main consequence
of these investigations is that ‘waves of probability’ can be produced in the
general situation (including macroscopic systems) owed to a combination of
a few preparation contexts In the two slit experiment we consider the screen
with slits as a part of an ensemble-preparation device and the screen with
photoemulsion as a measurement device.. Thus such ‘waves’ are not directly
related to some wave features of objects participating in experiments. More-
over, our investigation demonstrated that in some experiments there can
be created other types of probabilistic waves, namely, hyperbolic waves of
probability.
In particular, our contextual probabilistic investigations demonstrated
that contextual complementarity, wave–particle dualism, is not rigidly cou-
pled to microworld. Thus we can, in principle, perform experiments with
macro systems that would demonstrate ‘wave–particle duality’, but not of
macro objects, but contexts.
We recognize that at the moment there is one (and there seems to be
just one) argument supporting individual complementarity, namely, Bell’s
inequalitySome people use this framework to support quantum nonlocality.
In fact, Bell started the inequality story by having a definite aim, namely, to
show that every realist theory should be nonlocal. Realism was self-evident
for him as a consequence of precise anti-correlations. This original view-
point of Bell practically could not be recognized in modern publications in
that locality and realism are unified in ‘local realism’. I would like to thank
Shelly Goldstein who tried many times to explain to me Bell’s original view-
point.. However, I have great doubts that the experimental violation of Bell’s
inequality can be interpreted as an argument against the possibility of on-
tic HV-description (or even against realism) or locality, see, e.g., my book
[117]. In [117] it was demonstrated that by taking into account contextual
dependence of probabilities we immediately understand that Bell’s inequality
should be modified.
We remark that the possibility of the (99)-description implies that ‘quan-
tum randomness’ does not differ essentially from ‘classical randomness’. Of
course, this contradicts the orthodox quantum views of randomness as fun-

168
damental or irreducible randomness, see, e.g., [201]. Unfortunately I could
not understand the latter ideas. Instead of fundamental irreducible quantum
randomness I prefer to consider the readily understandable theory of context
depending probabilities, see section 2.
The main lesson of our consideration is that, as was already underlined
by Bell, [27], all ‘no-go’ arguments against the possibility of creating causal
deterministic models describing the underlining physical processes that pro-
duce quantum randomness are not, in fact, totally justified. Moreover, the
same critique can be applied to Bell’s arguments against local causal models.
However, we should not be rigidly coupled to the old fashioned Newtonian
continuous model of space–time.
In particular, we can study discrete causal models. Such modeling is
strongly supported by the whole quantum story. From the beginning ‘quan-
tum’ was really quantum — there were Planck and Einstein’s observations
on the appearance of the discrete energy structure instead of the continuous
real structure. In our papers with Jaroslav Volovich [121] we considered the
diffraction of charged balls (described by the discrete version of Newtonian
mechanics) on a screen with a slit. We also considered an analogue of the two
slit experiment. Suddenly one morning we saw on the computer’s screen in-
terference pictures which were very similar to ordinary quantum interference
images. The phenomenon is statistically stable — the picture is not changed
when the number of trajectories varies in a few millions (the total number
of trajectories in our numerical experiment was 50 millions). However, the
distribution of interference rings depends crucially on the relation between
discretizations of time and space parameters. Our discrete-time investiga-
tions raised the following questions:
a). Is physical space adequately described by the continuous real mathe-
matical model?
b). Was Newtonian mechanics just an approximation to discrete mechan-
ics? Are differential equations just approximations to difference equations?
1.2. Romantic interpretation.
Finally, we ask: ‘Why is the realist interpretation not so popular in the
quantum community?’ The common opinion is the following: this is the
direct consequence of experiments with elementary particles.
I do not think so. Of course, interference experiments with massive par-
ticles should induce reconsideration of the methods of classical statistical
mechanics. Discreteness of energy levels should induce reconsideration of the
classical continuous real model of physics. However, such reconsiderations
would be merely mathematical. And it seems (at least to me) that they
were merely mathematical. However, there were no reasons to create a new

169
quantum philosophy based on Bohr’s principle of complementarity and the
individual interpretation of Heisenberg’s uncertainty relations. Nevertheless,
such a new philosophy was invented (only by Bohr). Moreover, it was recog-
nized as the philosophy of modern physics. Since this recognition all realist
models for experiments with elementary particles were more or less automat-
ically rejected (for example, Bohmian mechanics). Finally, Bell’s inequality
arguments were interpreted as they should be interpreted in the orthodox
quantum framework, despite very strong counter-arguments.
I suspect that the main reason for this rather strange situation in modern
physics is the great attraction of the romantic spirit of orthodox quantum
philosophy. All these non-real things, wave–particle dualism, collapse, nonlo-
cality, irreducible randomness, were attractive for some of the creators as well
as for further quantum generations. These are two totally different stories:
a) to discuss merely mathematical modifications of the real-continuous
model of classical statistical mechanics;
or
b) to declare the scientific revolution.
Thus the orthodox Copenhagen interpretation was a kind of romantic
stage in the development of physics. It is always not easy argue against
romanticism. Probably we need not do this. I hope that realism (as the
history of literature shows us) would (sooner or later) come to physics.

51 Contextual Viewpoint of Quantum Stochas-

tics
It is well known that the classical rule for the addition of probabilistic alter-
natives:
P = P1 + P 2 (101)
does not work in experiments with elementary particles. Instead of this rule
we have to use the quantum rule:
p
P = P1 + P2 + 2 P1 P2 cos θ. (102)

The classical rule for the addition of probabilistic alternatives is perturbed

by the so called interference term. The difference between the ‘classical’
and ‘quantum rules was (and is) the source of permanent discussions as well
as various misunderstandings and mystifications, see, e.g., [21], [22], [37],
[64], [89], [90] for general references. We just note that the appearance of
the interference term was the source of the wave-viewpoint of the theory of
elementary particles. At least the notion of superposition of quantum states

170
was proposed as an attempt to explain the appearance of a new probabilistic
calculus in the two slit experiment, see, for example, Dirac’s book [64] on the
historical analysis of the origin of quantum formalism.
In particular, this induced the viewpoint that there are some special
‘quantum’ probabilities that differ essentially from ordinary ‘classical’ proba-
bilities. We also remark that the orthodox Copenhagen interpretation of the
quantum formalism (especially the principle of complementarity) is just an
attempt to explain (102) without appealing to mysterious ‘quantum probabil-
ities’. To escape the use of a new probabilistic calculus we can suppose that,
e.g., electron participating in the two slit experiment is in the superposition
of passing through both slits.
The role of an experimental context — a complex of physical conditions
— in a quantum measurement was discussed in the details already at the first
stage of development of quantum theory. Bohr always pointed out that the
experimental arrangement plays the crucial role in quantum theory. Heisen-
berg also paid much attention to the role of an experimental context (includ-
ing the derivation of the uncertainty principle), see [89]; see also L. Ballentine
[22], p. 367:
“The desirability of an analysis of the measurement process by means of quan-
tum theory was perhaps first indicated through Bohr’s insistence that the ‘whole
experimental arrangement’, object plus apparatus, must be taken into account in
order to specify well defined conditions for an application of the theory. However,
he never carried this program to its logical conclusion; the description of both the
object and measuring apparatus by the formalism of quantum theory.”
It seems that Bohr’s context is a kind of geometry, an experimental ge-
ometry. The only difference between ordinary geometry and contextual ge-
ometry is that the first is an individual, point, model of physical reality and
the second is a statistical model of physical reality.
At the first stage of the development of the quantum formalism contex-
tualism was merely discussed in the general philosophical framework. Of
course, such a discussion played a great role in the creation of the founda-
tions of quantum mechanics. In particular, Bohr’s complementarity principle
was created on the basis of contextual discussions, see section 1 of this chap-
ter. Unfortunately, contextualism did not provide a mathematical model
that could explain exotic features of the quantum experimental statistics. In
particular, there was not found a contextual (purely classical) probabilistic
derivation of quantum probabilistic rule (102). This inability to obtain inter-
ference for quantum particles considered as classical–realistic objects — but
moving in varying contextual geometries — induced a ‘romantic viewpoint’
of quantum systems: wave features, absence of trajectories,..., see section 1.

171
 There are many reasons why contextualism was not successful in deriv-
ing quantum statistics. One of the problems was of a purely mathematical
character. The standard probabilistic formalism based on Kolmogorov’s ax-
iomatics [126] of 1933 was the fixed context formalism. This conventional
probabilistic formalism does not provide rules for operating with probabili-
ties calculated for different contexts. However, in quantum theory we have to
operate with statistical data obtained for different complexes of physical con-
ditions, contexts. In fact, this context dependence upon probabilities as the
origin of the superposition principle has already been discussed by Heisen-
berg [89]; unfortunately, only in a quite general and a rather philosophical
framework. Later on the contextual dependence of quantum probabilities was
intensively investigated from various viewpoints. However, a simple contex-
tual derivation of the quantum probabilistic rule (102) was never presented.
Typically, contextual papers explain that contextual transitions can violate
the classical probabilistic rule (101). However, it was not clear why we have
the interference rule (102) in the quantum formalism and why we use the
Hilbert space calculus of quantum probabilities.
Such a classical probabilistic derivation of (102) was found by author,
see [123], [124]. In these papers we used von Mises’ frequency probability
model [143]. R. von Mises frequency approach induces quite complicated
manipulations with frequencies [143], [117] that are, in fact, irrelevant to
our physical considerations. Therefore in this chapter we proceed without
to go to the frequency level. On the other hand, frequency analysis gives
more detailed understanding of the contextual approach. We shall consider
the contextual probabilistic approach to quantum theory in the frequency
framework in chapter 10.

172
 2.1. Role of physical conditions.
In fact, probabilities in the transformation (102) in quantum experiments
are determined by at least three different contexts, S, S1 , S2 . In fact, (S1 , S2 )
is a kind of statistical coordinate system and (102) is an expansion of the
statistical variable, S-probability, with respect to this coordinate system. We
illustrate this situation with the following fundamental example.
Example 2.1. (Two slit experiment) This is the simplest experiment for
demonstrating the interference of light. There is a point source of light O
and two screens M and M 0 . The screen M contains two slits, numbered by
j = 1, 2. Light passes through M (through the slits) and finally reaches the
registration screen M 0 where we observe the interference fringes.
The wave explanation of the existence of interference fringes is well known:
the light reaching M 0 can travel by one of two routes – either through the
first slit or through the second slit. But the distances travelled by the light
waves following these two paths are not equal and the light waves do not
generally arrive at the screen ‘in step’ with each other.
On the other hand, O is a source of quantum particles, photons. To
exclude the interaction between photons in a beam we perform the experi-
ment with very weak light, so that at any time there is only one photon in
the region between O and M. The screen M 0 is replaced by a photographic
plate or film (also denoted by M 0 ). Individual spots appear on M 0 randomly.
However, there appear standard interference fringes after a sufficiently long
exposure. By this experiment we can compute the probability distribution
of spots on M 0 . Then we perform the analogous experiment in which one of
the slits is closed. We compute the corresponding probability distributions.
Because each photon goes through one of the two slits it seems that these
probabilities should be connected by the standard formula for the addition
of probabilities for alternatives. However, this formula is not valid for the
experimental probabilistic distributions.
Similar experiment were performed not only with photons, but also with
massive particles, e.g., electrons. The result was the same — experimental
statistical data violated the classical formula for the addition of probabilities.
We underline again the main distinguishing feature of statistical data
obtained in these three experiments. By combining through (101) the data
obtained in experiments in which only one of slits is open we do not get the
probability distribution for data obtained in the experiment in which both
slits are open. On the other hand, we never observe a particle which passes
through both slits simultaneously. It is observed passing either through the
first or through the second slit. There is no the direct observation of particle
splitting. Because each particle passes only through one of the slits, we have
the standard case of alternatives. Thus we should use the conventional rule

173
(101) for the addition of the probabilities of alternatives. This disagreement
between experimental statistical data and the rule of conventional probability
theory looks like some kind of paradox. The traditional solution of this
paradox is the use of the wave model for elementary particles, see Dirac [64]:
“This will present to us the difficulty of the conflict between the wave and
corpuscular theories of light in an acute form. Corresponding to the description
that we had in the case of the polarization, we must now describe the photon as
going partly into each of the two components into which the incident beam is split.”
The crucial point is that in the two slit experiment the rule (102) is
induced by combining the statistical data obtained in three different exper-
iments: both slits are open; only the jth slit is open, j = 1, 2. We perform
detailed contextual analysis for the two slits experiment. We consider the
following complexes of physical conditions (contexts):
S = both slits are open,
Sj =only the jth slit is open, j = 1, 2.
In fact, probabilities in (102) are related to these three contexts. Thus
P = PS (E) and Pj = PS→Sj PSj (E), j = 1, 2.
Here we use various context indexes. The PS (E), PSj (E) denote the prob-
abilities of an event E with respect to various contexts. The coefficients
PS→Sj , j = 1, 2, have another meaning. In general these are not probabili-
ties of Sj with respect to the context S (besides some very special, classical,
situations) because the context Sj in general is not an event for the context
S. The PS→Sj , j = 1, 2, are kinds of splitting probabilities. These are pro-
portion coefficients in the splitting of the ensemble S, prepared on the basis
of the complex of physical conditions S, into ensembles Sj prepared on the
basis of the complexes of physical conditions Sj :

Nj
PS→Sj = , (103)
N
where N is the number of elements in S and Nj is the number of elements
in Sj , see chapter 10 for the details. We remark (and it is important for
our further considerations) that we have the following condition of statistical
alternatives:
PS→S1 + PS→S2 = 1. (104)
The condition of statistical alternatives has the following meaning: the total
number of particles which arrive at the registration screen when both slits are
open is equal (on average) to the sum of the corresponding numbers when
only one of the slits is open. So by closing, e.g., the first slit we do not
change the number of particles which pass the second slit (on average). In

174
fact, (104) gives the right description of the alternative situation in the two
slit experiment. It is not related to alternative passing through the slits by a
particle in the experiment when both slits are open. This equation describes
alternative sharing of particles between two preparation procedures: the jth
slit is open, j = 1, 2.
However, the splitting probabilities PS→Sj would not play so important a
role in our considerations. The crucial role will be played by the contextual
probabilities PS (E), PSj (E).
The conventional probability theory says, see, e.g., [117], that we have:

P (E) = P (S1 )P (E/S1 ) + P (S2 )P (E/S2 ). (105)

if S1 ∩ S2 = ∅ and S1 ∪ S2 = S. This is the well known formula of to-

tal probability. In many considerations (including the works of the fathers
of quantum mechanics, see, e.g., Dirac [64]) people set P = P (E) and
Pj = P (S1 )P (E/S1 ). Finally, they obtain the contradiction between the
conventional probabilistic rule (101) (which in fact coincides with (105)) and
statistical data obtained in the interference experiments and described by
the quantum rule (102).
We would like to discuss the physical and mathematical assumptions used
for the derivation of (105). The main physical assumption is that this formula
is derived for one fixed context S. In the mathematical formalism this is a
single fixed Kolmogorov probability space; see my book [117] on extended
discussions for the role of the choice of Kolmogorov’s probability space in
quantum measurements. To be precise, we have to write this formula as

PS (E) = PS (S1 )PS (E/S1 ) + PS (S2 )PS (E/S2 ). (106)

It is also important that the contexts S1 , S2 can be realized as elements of

the field of events corresponding to the context S. Thus we would obtain the
contradiction between the classical rule (101) and the quantum rule (102)
only if we assume that the splitting probabilities PS→Sj can be interpreted as
PS -probabilities, PS→Sj = PS (Sj ), and the contextual probabilities PSj (E) as
conditional probabilities with respect to the context S, PSj (E) = PS (E/Sj ).
Here the contextual probabilities are given by Bayes’ formula (the additional
postulate of Kolmogorov’s probability theory):

PS (E ∩ Sj )
PSj (E) = .
PS (Sj )

In general there are no reasons for assuming that new complexes of conditions
Sj are ‘so nice’ that new probability distributions are given by the Bayes’

175
formula. Thus, in general, the formula of total probability can be violated
when we combine statistical data obtained for a few distinct contexts. In
particular, this formula does not hold true in statistical experiments with
elementary particles. The right hand side of this formula is perturbed by
the interference term. This explanation is well known in the contextual
community.
Remark 2.1. We can discuss the same experiment in the standard frame-
work of preparation/measurement procedures. The context S produced an
ensemble of particles S which passed through the screen with slits when both
slits are open. A measurement procedure is the measurement of the position
of a particle on the registration screen. In the same way the context Sj pro-
duced an ensemble of particles Sj which passed through the screen with slits
when only the jth slit was open.
Remark 2.2. I recently discovered that, in fact, there is no contradiction
between Kolmogorov’s approach [126] and the contextual approach I would
like to thank Professor A. Shiryaev (a former student of A.N. Kolmogorov)
who explained this to me.. It is a mysterious fact that a few generations of
readers (including the author of this book) did not pay attention to section
2 of Kolmogorov’s book [126]. There Kolmogorov said directly that all prob-
abilities of events are related to concrete complexes of physical conditions.
2.2. Interference term as the measure of statistical deviations.
The following simple consideration gives us the derivation of the quantum
probabilistic transformation (102) in the classical probabilistic framework.
Let S and Sj , j = 1, 2, be three different complexes of conditions. We
consider the transformation of probabilities induced by transitions from one
complex of conditions to others:

S → S1 and S → S2 . (107)

We start by introducing the splitting probabilities PS→Sj . These are pro-

portional coefficients for numbers of physical systems obtained after prepara-
tions under the complexes of physical conditions S and Sj . If (starting with
the same number of particles) we obtain N and Nj systems after S and Sj
preparations, respectively, then the PS→Sj are defined by (103). We assume
that splitting probabilities satisfy the statistical alternative equation (104).
This is a natural condition: splitting (107) of the context S induces just shar-
ing of a physical systems produced by a source. We have already discussed
this sharing in the two slit experiment. In neutron interferometry we have
the same situation for the sharing of particles coming to the detectors when
both paths are open and when just one of the paths is open.

176
 We introduce the measure of statistical perturbations δ induced by context
transitions:
δ(S → S1 , S2 ; E) = PS→S1 [PS (E) − PS1 (E)] + PS→S2 [PS (E) − PS2 (E)].
This quantity describes the deformation of probability distribution PS due to
context transitions. By using statistical alternative equation (104) we obtain:
PS (E) = PS (E)PS→S1 + PS (E)PS→S2 .
Thus we obtain:
PS (E) = PS→S1 PS1 (E) + PS→S2 PS2 (E) + δ(S → S1 , S2 ; E). (108)
Transformation (108) is the most general form of probabilistic transforma-
tions induced by a context transitions. There is the correspondence principle
between context unstable and context stable (‘classical’) transformations:
If δ(S → S1 , S2 ; E) → 0 then the contextual probabilistic transformation
(108) coincides (in the limit) with the conventional formula of total proba-
bility.
The perturbation term δ(S → S1 , S2 ; E) depends on the absolute mag-
nitudes of the probabilities. It would be natural to introduce a normalized
coefficient of the context transition
δ(S → S1 , S2 ; E)
λ(S → S1 , S2 ; E) = p ,
2 PS1 /S PS1 (E)PS→S2 PS2 (E)
which gives the relative measure of statistical deviations induced by the tran-
sition from one complex of conditions, S, to others, Sj , j = 1, 2. The trans-
formation (108) can be written in the form:
X
PS (E) = PS→Sj PSj (E)
j=1,2
p
+2 PS→S1 PS1 (E)PS→S2 PS2 (E)λ(S → S1 , S2 ; E) . (109)
In fact, there are two possibilities:
1). |λ(S → S1 , S2 ; E)| ≤ 1;
2). |λ(S → S1 , S2 ; E)| ≥ 1.
In both cases it is convenient to introduce a new context transition pa-
rameter θ = θ(S → S1 , S2 ; E) and represent the context transition coefficient
in the form:
λ(S → S1 , S2 ; E) = cos θ(S → S1 , S2 ; E), θ ∈ [0, π];

177
and
λ(S → S1 , S2 ; E) = ± cosh θ(S → S1 , S2 ; E), θ ∈ [0, ∞),
respectively.
We have two types of probabilistic transformations induced by the tran-
sition from one complex of conditions to another:
X
PS (E) = PS→Sj PSj (E)
j=1,2

p
+2 PS→S1 PS1 (E)PS→S2 PS2 (E) cos θ(S → S1 , S2 ; E) . (110)
X
PS (E) = PS→Sj PSj (E)
j=1,2
p
±2 PS→S1 PS1 (E)PS→S2 PS2 (E) cosh θ(S → S1 , S2 ; E) . (111)
We derived the quantum probabilistic rule (102) in the classical proba-
bilistic framework (in particular, without any reference to superposition of
states) by taking into account the context dependence of the probabilities.
Relatively large statistical deviations are described by the transformation
(111). Such transformations do not appear in the conventional formalism
of quantum mechanics. In principle, they could be described by so called
hyperbolic quantum mechanics, see chapter 10.
Conclusion. For each fixed context (experimental arrangement) we have
classical statistics. Context transition can induce interference perturbation
of the conventional rule for the addition of probabilistic alternatives.
2.3. Linear algebra for probabilities, complex amplitudes.
One of the main distinguishing features of quantum theory is the Hilbert
space calculus for probabilistic amplitudes. As we have already discussed,
this calculus is typically associated with the wavelike (superposition) features
of quantum particles. We shall show that, in fact, the Hilbert space represen-
tation of probabilities was merely a mathematical discovery. Of course, this
discovery simplifies calculations essentially. However, this is pure mathemat-
ics; physics is related merely to the derivation of the quantum interference
rule (102).
The crucial (physical!) point was the derivation (at the beginning purely
experimental) of the transformation (102) connecting probabilities with re-
spect to three different contexts. In fact, linear algebra (mathematics!) can
be easily derived from this transformation. Everybody familiar with the ele-
mentary geometry will see that (102) is just the well known cosine-theorem.

178
This is the rule to find the third side in a triangle if we know lengths of two
other sides and the angle θ between them:

c2 = a2 + b2 − 2ab cos θ .

or if we want to have "+" in front of the cosine we use the so called parallel-
ogram law:
c2 = a2 + b2 + 2ab cos θ . (112)
Here c is the diagonal of the parallelogram with sides a and b and the angle θ
between these sides. Of course, the parallelogram law is just the law of linear
(two-dimensional Hilbert space) algebra: for finding the length c of the sum
c of vectors a and b having lengths a and b and an angle θ between them.
We also can introduce complex waves by using the following elementary
formula:
a2 + b2 + 2ab cos θ = |a + beiθ |2 . (113)
Thus the context transitions S → Sj , j = 1, 2, can be described by the wave:
p p
ϕ = PS→S1 PS1 (E) + PS→S2 PS2 (E)eiθ(S→S1 ,S2 ;E) .

2.4. Classical probabilistic derivation of the superposition prin-

ciple.
We shall study in more detail the possibility of contextual (purely clas-
sical) derivation of the superposition principle for complex probability am-
plitudes, ‘waves’, in the two slit experiment. We consider one dimensional
model. It could be obtained by considering the distribution of particles on
one fixed straight line, a very thin strip. It is supposed that the source of
particles is symmetric with respect to the slits and the straight line on the
registration screen passes through the center of the screen. This geometry
implies that PS→Sj = 1/2, j = 1, 2. By the symbol Ex , x ∈ R, is denoted the
event of the registration of a particle at the point x of the straight line. We
set p(x) = PS (Ex ) and pj (x) = PSj (Ex ), j = 1, 2, where contexts S and Sj
were defined in Example 2.1. By using (110) we obtain:
1 p
p(x) = [p1 (x) + p2 (x) + 2 p1 (x)p2 (x) cos θ(x)].
2
By using (113) we represent this probability as the square of a complex
amplitude, p(x) = |φ(x)|2 , where
1 p p
φ(x) = √ (eiθ1 (x) p1 (x) + eiθ2 (x) p2 (x)) (114)
2

179
and phases θj (x) are chosen in such a way that the phase shift θ1 (x)−θ2 (x) =
θ(x). We also introduce complex amplitudes for probabilities
1
q
pj (x) : φj (x) = √ eiθj (x) pj (x).
2
Here pj (x) = |φj (x)|2 . The complex amplitudes are said to be wave functions:
φ(x) is the wave function for the context ‘both slits are open’; φj (x) is the
wave function for the context the ‘jth slit is open’, j = 1, 2.
Let us set ξ(x) = θ(x)/h, where h > 0 is some scaling factor. We have:
1  iξ1 (x) p iξ2 (x) p

φ(x) = √ e h p1 (x) + e h p2 (x)
2
and
1 iξj (x)
q
φj (x) = √ e h pj (x).
2
By choosing h equal to the Planck constant we obtain a quantum-like repre-
sentation of probabilities. We recall that we have not used any kind of wave
arguments. The superposition rule (114) was obtained in purely classical
probabilistic (but contextual!) framework.
Suppose now that ξ depends linearly on x :
pj x px
ξj (x) = , ξ(x) = ,
h h
where p = p1 − p2 . Under such an assumption we shall obtain interference of
two ‘free waves’ corresponding to momenta p1 and p2 . Of course, this linearity
could not be extracted from our general probabilistic considerations. This is
a consequence of the concrete geometry of the experiment.
2.5. The coefficient of context transition as the measure of in-
compatibility of physical observables.
We now consider the relation between the coefficient of context transition
(the measure of statistical deviations induced by the change of complex of
physical conditions) and incompatibility of physical observables in quantum
mechanics (noncommutativity of the corresponding operators). As every-
where in this paper, we consider dichotomous observables. Each event E
generates the dichotomous variable A : A = a1 if E occurs and A = a2 if E
does not occur. Values a1 and a2 do not play any role in our considerations;
in principle, we can consider the case a1 = 0 and a2 = 1.
Definition 2.1. A physical observable A is incompatible with a pair
S1 , S2 of contexts if there exists a context S such that
δ(S → S1 , S2 ; A = ai ) 6= 0.

180
 In such a case a transition from the complex of physical conditions S to
complexes Sj induces non-negligible statistical deviations for A-measurements.
The measurement of the observable A under the complex of conditions S dif-
fers from the measurement of the same observable A under the complex of
conditions Sj , j = 1, 2.We need to consider two complexes S1 and S2 be-
cause we would like to consider another dichotomous variable B connected
with these contexts.
We shall demonstrate that the incompatibility of physical observables in
quantum mechanics is just a particular case of contextual incompatibility.
Let H be the two-dimensional Hilbert space. Rays of this space represent
some class of complexes of physical conditions. Let the dichotomous variable
A be represented by a self-adjoint operator (a symmetric matrix) A. b We
remark that we can associate with any physical observable A two complexes
of conditions S1a , S2a , namely, contexts represented by eigenvectors φa1 , φa2 of
b The context S a describes the filter (selection) with respect to the value
A. j
A = aj .
Let us consider another dichotomous physical observable B = b1 , b2 . It is
represented by a self-adjoint operator B b with eigenvectors φb1 , φb2 . These eigen-
vectors represent contexts S1b , S2b (filtrations corresponding to values B = b1
and B = b2 , respectively).
Theorem 2.1. Quantum physical observables A and B are incompat-
ible (i.e., corresponding operators do not commute) iff the observable A is
incompatible with the contexts Sjb or vice versa:

δ(S → S1b , S2b ; A = ai ) 6= 0 or δ(S → S1a , S2a ; B = bi ) 6= 0.

Proof. Let S be an arbitrary quantum context. Thus it can be repre-

sented by a normalized vector φ ∈ H. We have:

δ(S → S1b , S2b ; A = ai ) =

PS→S1 [PS (A = ai ) − PS1 (A = ai )] + PS→S2 [PS (A = ai ) − PS2 (A = ai )]

= |(φ, φb1 )|2 (|(φ, φai )|2 − |(φai , φb1 )|2 ) + |(φ, φb2 )|2 (|(φ, φai )|2 − |(φai , φb2 )|2 ).
We have (φ, φbj ) = kj eiξj , (φbj , φai ) = kji eiξji , where kj , kij ≥ 0. We obtain:

δ(S → S1b , S2b ; A = ai ) = 2k1 k2 k1i k2i cos θi ,

where θi = ξ2 − ξ1 + ξ2i − ξ1i .

a). Let [A, b = 0. Then k12 = k21 = 0. Hence δ(S → S1 , S b ; A = ai ) = 0.
b B]
2

181
 b). Let [A, b 6= 0. Then k12 , k21 6= 0. Let k1 , k2 > 0 be arbitrary constants
b B]
such that k1 + k22 = 1. We choose a context S that is described by the state:
2

p p
φ = k1 eiξ21 φb1 + k2 eiξ11 φb2 .

Here θ = 0, and hence

δ(S → S1b , S2b ; A = ai ) = 2k1 k2 ki1 ki2 > 0.

52 Law of Statistical Balance in Nature

A. Zeilinger [201] remarked:
“It so happened that almost all relativistic equations which appear in Einstein’s
publication of 1905 were known already before, mainly through Lorentz, Fitzgerald
and Poincare — simply as an attempt to interpret experimental data quantitatively.
But only Einstein created the conceptual foundations, from which, together with
the constancy of the velocity of light, the equations of the relativity arise. He did
this by introducing the principle of relativity...”
Zeilinger was looking for a new quantum paradigm. He correctly under-
lined that the situation in quantum theory, especially the large diversity of
interpretations, is not so natural. That in the opposite of, e.g., the the-
ory of relativity, there is no quantum analogue of the fundamental principle.
Following Zeilinger [201] we are looking for an analogue of the fundamental
principle which would play the role of Einstein’s relativity principle.
Ideas which are similar to Zeilinger’s idea of finding a ‘quantum fun-
damental principle’ have been intensively discussed by many authors. In
particular, extended discussions took place at the last Växjö Conference on
foundations of quantum theory, see [121]. I would like to mention the papers
of C. Fuchs, L. Hardy and J. Summhammer in [121].
Here I propose such a quantum fundamental principle. We start again
with the remark that the quantum formalism is a statistical theory. A theory
which deals with probabilities of eventsHowever, as we have already dis-
cussed, the existence of such a probabilistic description does not imply the
impossibility óf a realist ontic description.. Therefore a fundamental quan-
tum principle (if it exists at all) should be a statement about probabilities.
We are looking for a fundamental statistical principle which would induce
the quantum formalism.
3.1. General statistical measurement theory.
We start with consideration of the general statistical measurement theory.
This theory is based on the following experimental fact:

182
 (CP) All observed probabilities depend on complexes of experimental phys-
ical conditions, contexts.
This principle of contextuality of probabilities is very general. It is also
more or less evident (at least after it has been formulated)The notion of
context is closely related to a so called preparation procedure, see, e.g., [21].
However, we prefer context, since context need not be prepared by somebody.
It can be prepared by Nature.. It seems that it would be impossible to derive
some concrete theoretical results starting from only this principle. But it is
not so. I was surprised by myself when it was found [123], [124], see also
section 2 of this chapter, that many important consequences can be derived
by using only this principle, (CP). We just formulate consequences of the
principle (CP).
We are interested in transformations of probabilities for physical observ-
ables induced by transitions from one context to another. In the general
statistical measurement theory a context plays the role of a system of coor-
dinates (statistical coordinates). Thus we are looking for transformations of
statistical coordinates.
To escape technical complications we consider two physical observables
A and B that are dichotomous: A = a1 , a2 and B = b1 , b2 , see [124], p. 9967
for the general case of discrete multi-valued observables. We consider some
complex of physical conditions, a context, S (‘a system of coordinates’).
By using S we can prepare a large ensemble S of physical systems. By
performing the A-measurement over elements of S we obtain probabilities (a
kind of vector for a ‘context-coordinate system’):
paj = PS (A = aj ), j = 1, 2.
We now consider contexts Sib , i = 1, 2, that realize filtrations with respect
to the values B = bi of the B-observable. Let Sib , i = 1, 2, be correspond-
ing ensembles of physical systems. By performing the A-measurement over
elements of Sib we obtain the probabilities
a/b
pij = PSib (A = aj ), i, j = 1, 2.
We would like to predict the probabilities paj for the A-measurement for an
a/b
arbitrary context S by knowing the probabilities pij of the A-measurement
for the contexts Sib , i = 1, 2, corresponding to fixed values B = bi of the
B-observable.
We have found [123], [124], see section 2, that such a prediction rule can
always be represented in the form of the transformation:
q
a a/b a/b a/b a/b
pj = p1 p1j + p2 p2j + 2 p1 p2 p1j p2j λj , (115)

183
j = 1, 2. Here pi = PS→Sib , i = 1, 2, are splitting probabilities, see section 2.
They are computed in the following way. By applying the filtration B = bi to
the ensemble S (prepared on the basis of the complex of physical conditions
S) we obtain the ensembles Sib . Then

|Sib |
pi ≈ ,
|S|

i = 1, 2, when N → ∞, where N is the number of elements in S. We use the

symbol |O| to denote the number of elements in the set O.
In the transformation (115) the coefficients λj = λ(S → Sib ; A = aj ) are
context transition parameters. They give the measure of perturbations of
probabilities for the observable A = aj induced by the transitions from the
context S to the contexts Sib , i = 1, 2.
Remark 3.1. (Stochasticity of the matrix of transition probabilities).
We always have
p11 + p12 = 1 and p21 + p22 = 1, (116)
since pi1 + pi2 = PSib (A = a1 ) + PSib (A = a2 ) and, for probabilities corre-
sponding to the fixed complex of physical conditions (in our case Sib ), we use
the standard rule for the addition of probabilities of alternatives (in our case
a/b
A = a1 or A = a2 ). A matrix P = (pij ) satisfying (116) is called a stochastic
matrix.
In [123], [124], see section 2, we classified statistical measurement theo-
ries with respect to magnitudes of the context transition parameters λ(S →
Sib ; A = aj ) :
a) trigonometric theory: |λj | ≤ 1, j = 1, 2, so these parameters can be
represented as λj = cos θj , θj ∈ [0, 2π];
b) hyperbolic theory: |λj | > 1, j = 1, 2, so λj = ± cosh θj , θj ∈
(−∞, +∞);
c) hyper-trigonometric theories: |λ1 | ≤ 1 and |λ2 | > 1 or |λ1 | > 1 and
|λ2 | ≤ 1, so λ1 = cos θ1 , λ2 = ± cosh θ2 or λ1 = ± cosh θ1 , λ2 = cos θ2 .
We are interested in trigonometric statistical theories. Here we study
context transitions which produce relatively small perturbations of probabil-
ities:
|λ(S → S1b , S2b ; A = aj )| ≤ 1, (117)
j = 1, 2. In particular, we can consider a statistical measurement theory that
operates with classes of contexts and observables such that all the coefficients
λj = 0. Such statistical theories of measurement we call classical statistical
theories. In classical theories the general transformation (115) of probabilities

184
corresponding to context transitions is reduced to the well known formula of
total probability
a/b a/b
paj = p1 p1j + p2 p2j . (118)
Remark 3.2. (The formula of total probability in the conventional prob-
abilistic formalism). Let (Ω, F, P) be Kolmogorov’s probability space [126],
see e.g., [117]. Here Ω is the space of elementary events, F is the σ algebra
of events and P-probability measure. Physical observables A = a1 , a2 and
B = b1 , b2 are represented by random variables A(ω), B(ω), ω ∈ Ω. Condi-
tional probability is defined by Bayes’ formula:
P(A = aj , B = bi )
P(A = aj /B = bi ) = .
P(B = bi )
By using additivity of probability we obtain the formula of total probability:

P(A = aj ) = P(B = b1 )P(A = aj /B = b1 )

+P(B = b2 )P(A = aj /B = b2 ) (119)

a/b
Here paj = P(A = aj ), pi = P(B = bi ), pij = P(A = aj /B = bi ). Here the
complex of physical conditions Sib can be mathematically represented as an
element of the σ-algebra of events F − Sib = Hi = {ω ∈ Ω : B(ω) = bi } ∈ F.
In statistics events Sib are considered as statistical hypotheses. Formula
(119) is used for the prediction of the probability of the event Ej = {ω ∈ Ω :
A(ω) = aj } if we know the probabilities of this event under the statistical
hypothesis Hi − P(Ej /Hi ) :

P(Ej ) = P(H1 )P(Ej /H1 ) + P(H2 )P(Ej /H2 ). (120)

However, if perturbations of probabilities corresponding to context transi-

tions
S → Sib
are relatively large, namely, λi 6= 0, we can not use the standard Bayes’ rule
to predict the probability PS (A = aj ) of the event Ej = {A = aj } under the
context S on the basis of probabilities PSib (A = aj ) of the event Ej under
contexts Sib .
Remark 3.3. (Contextual statistics). It seems natural from the general
statistical viewpoint to develop an analogue of Bayesian statistical analysis
(see Remark 3.2) for a statistical hypothesis Hi which could not be repre-
sented mathematically as elements of one fixed σ−algebra F of events. I
tried to find such a statistical formalism in the literature; unfortunately, I
could not ... .

185
 In particular, in a trigonometric statistical theory, see (117), the standard
prediction rule (118) is modified:
q
a/b a/b a/b a/b
paj = p1 p1j + p2 p2j + 2 p1 p2 p1j p2j cos θj . (121)

Here the ‘phases’ θj = θ(S → Sib ; A = aj ) have a purely probabilistic mean-

ing. These are simply convenient reparametrizations (see later considera-
tions) of the perturbation probabilistic coefficients λj . The reader has already
recognized in (121) the form of the quantum rule for interference of proba-
bilities. Quantum statistical theory is simply one of trigonometric statistical
theories of measurement.
However, the transformation (121) is too general for conventional quan-
tum theory (the Dirac–von Neumann formalism). If you perform calculations
in the Hilbert space of quantum states, you will see that the quantum trans-
formations (121) are characterized by the very special choice of matrices
a/b
P = (pij ) of transition probabilities. Matrices produced by the quantum
formalism are so called double stochastic matrices.

186
 3.2. The principle of statistical balance in nature.
As we have seen, we can not derive quantum theory by only using the prin-
ciple of contextuality of probabilities (CP). Of course, the quantum theory
of measurement is one of the statistical theories of measurement. However,
we have to find additional conditions that specify quantum statistical the-
ory amongst all possible statistical theories of measurement. So we have to
find some specific quantum principles — additional to the general statistical
principle (CP). We already know that quantum statistical theory belongs to
the class of trigonometric theories. These are theories describing classes of
contexts and physical observables such that transitions from one context to
another produce relatively small perturbations of probabilities, |λi | ≤ 1.
We can formulate this as a principle of relatively small perturbations,
(SP).
However, (SP) is not, in fact, a physical principle. This is the principle
on the human investigation of nature. We would like to consider theories
giving as small as possible errors of prediction. In principle we can study
some measurements over elementary particles described by the hyperbolic or
hyper-trigonometric statistical theory. It would simply mean that we consider
some observables which are essentially more sensible to context transitions
than quantum observables (we shall discuss this question later in more detail).
In fact, the crucial quantum point is double stochasticity of the matrix of
transition probabilities. The fundamental problem of quantum theory is not
interference of probabilities, but:
Why do we have for quantum observables and contexts not only stochas-
ticity condition (116) but also double stochasticity condition?
We recall that the condition of double stochasticity has the form:

p11 + p21 = 1 and p12 + p22 = 1, (122)

As we have seen, stochasticity is easily explained on the basis of general

probabilistic arguments. But double stochasticity? We have:

PS1b (A = a1 ) + PS2b (A = a1 ) = 1, (123)

PS1b (A = a2 ) + PS2b (A = a2 ) = 1. (124)

But why? I think that equations (123), (124) should be interpreted as sta-
tistical balance equations. If B = b1 preparation produces too much A = a1
property, then B = b2 preparation should produce just a little of the A = a1
property: if PS1b (A = a1 ) is large then PS2b (A = a1 ) must be small. And vice
versa.

187
 We underline that we need not consider A (or B) as an objective property,
the property of an object-a physical system. The A (or B) is a statistical
property. Probabilities for its values are assigned to statistical ensembles
corresponding to various contexts.
Equations (123),(124) are simply the mathematical representation of the
great law of nature:
The law of statistical balance.
By this law in the process of ‘preparations’ (which occur randomly in
nature) for, e.g., B = b1 , b2 the balance between the values A = a1 and
A = a2 of the A-property can not be violated. If the law of statistical balance
be violated, e.g., in favour of the value A = a1 , then after some period the
value A = a1 would strongly dominate over the value A = a2 . Finally, in the
limit t → ∞ the value A = a2 will disappear. Thus A will be no longer a
physical observable, since it will take the single value A = a1 everywhere.
Quantum theory is a statistical measurement theory which is based on
the principle of statistical balance:
(SB) Creation of physical properties by ‘preparations’ which randomly oc-
cur in nature does not violate statistical balance of physical properties.
On the one hand, (SB) is the fundamental principle of quantum theory,
the special theory of statistical measurements. On the other hand, (SB) is the
law of nature. This is the very special feature of quantum theory. A theory
which describes the most fundamental ‘preparations’ of physical properties in
nature. Such ‘preparations’ are performed at the level of elementary particles.
The violation of the law of statistical balance at such a level for some property
A would imply that such a property should sooner or later disappear at the
microlevel and, as a consequence, at the macrolevel. There could be such
properties at the initial stage of the evolution of the universe. However, they
should disappear. Therefore at the present time we observe only physical
properties of elementary particles that follow the law of statistical balance.
Of course, the law of statistical balance in nature can be violated for
some contexts and (or) observables which are not of a fundamental charac-
ter. Moreover, we can find observables which would follow non-trigonometric
transformation laws under a transition from one context to another.
Conclusion: Quantum theory is a statistical theory of measuremen-
tWith relatively small perturbations of probabilities corresponding to context
transitions. based on the principle of statistical balance in nature.
Of course, we do not forget about Planck’s constant h. This is the con-
stant of action which can not be changed by varying the system of ‘statistical

188
coordinates’, the context (= complex of physical conditions). This is an ana-
logue of the constant velocity of light in Einstein’s theory.
We formulated quantum theory in the way similar to Einstein’s formula-
tion of general relativity:
a) The quantum theory of measurement is a theory about systems of sta-
tistical coordinates, contexts, and physical observables which are measured
in such systems. Numerically these statistical coordinates for observables are
given by the probabilities paj = PS (A = aj ).
b) This theory gives the possibility of changing statistical coordinates
by using the transformation (121). The trigonometric form of this transition
law is a consequence of quantum theory being concerned with relatively small
perturbations of statistical coordinates corresponding to transitions from one
system of coordinates to another system.
c) The fundamental principle of quantum theory is the principle of statis-
tical balance in nature. This principle implies that transition matrices corre-
sponding to transformations of statistical coordinates are double stochastic
matricesWe remark that we consider only dichotomous variables..

53 Experiments on Quantum-Like Behaviour

of the Mind
We describe cognitive experiments (based on interference of probabilities for
mental observables) which could verify a quantum-like structure of mental
measurements. In principle, such experiments could be performed in psy-
chology, and cognitive and social sciences.
We have already underlined on a few occasions that for us quantum the-
ory is merely a special theory of statistical averages. The main experimen-
tal consequence of special quantum probabilistic behaviour is interference of
probabilities of alternatives. As we have remarked, in classical statistical
physics the probability of the event C = A or B, where A and B are alter-
natives, is equal to the sum of the probabilities. In quantum physics there
appears an additional additive term, the interference term.
By using such a point of view of quantum theory we can use its methods
not only to describe measurements over elementary particles, but also over
other systems which could demonstrate quantum probabilistic behaviour.
We describe a mental interference experiment.mental interference exper-
iment
Let A = a1 , a2 and B = b1 , b2 be two dichotomous I-observables: a1 =
‘yes’, a2 = ‘no’, b1 = ‘yes’, b2 = ‘no’. They can be two different questions or

189
two different types of cognitive tasks. We prepare an ensemble S of cognitive
systems (e.g., human beings) having the same mental state. Then we perform
measurements of A over elements of the ensemble S and obtain ensemble
probabilities:
the number of results aj
paj = .
the total number of elements
So paj is the probability of obtaining the result aj under the measurement
over cognitive systems belonging to the ensemble S. In the same way we
find probabilities pbj for the B-observableWe pay attention to our need to
prepare a new ensemble S (in the same state) to perform the B-measurement.
We could not perform A and B measurements on the same system, since
the A-measurement perturbs the mental state essentially.. The next step
is to prepare two ensembles Sib , j = 1, 2, of cognitive systems having the
states corresponding to values of B = bj , j = 1, 2. The ensembles Sjb can
be prepared by using, e.g., filtrations with respect to values (e.g., answers)
B = bj , j = 1, 2. Hence, if we, e.g., ask the question B of a cognitive system
τ belonging to the ensemble Sjb , then we shall obtain the answer B = bj with
probability one.
We perform now the A-measurements for elements of the ensembles Sjb , j =
1, 2, and obtain the probabilities:
a/b the number of the result aj for the ensemble Sib
pij =
the total number of elements in Sib
a/b
For instance, the probability p12 is obtained as the frequency of the
answer A = a1 =yes in the ensemble of cognitive system that have answered
B = b2 = no.
Classical probability theory tells us that all these probabilities have to be
connected by the formula of total probability:
a/b a/b
paj = pb1 p1j + pb2 p2j ,
j = 1, 2. However, if the theory is quantum-like then we will obtain the
quantum formula of total probability:
q
a/b a/b a/b a/b
paj = pb1 p1j + pb2 p2j , +2 pb1 pb2 p1j p2j cos θj .
Here θj is the phase of the A-interference between the state of mind in
the ensemble S and the ensembles Sjb .
In the experiment on the quantum statistical test for mental measure-
ments we calculate
a/b a/b
paj − pb1 p1j + pb2 p2j
cos θj = q .
b b a/b a/b
2 p1 p2 p1j p2j

190
If cos θj 6= 0 then we would obtain the strongest argument in support of
quantum-like behaviour of cognitive systems. In this case, starting from (ex-
perimentally calculated) cos θj we can proceed to the Hilbert space formalism,
see chapter 10.

54 Experimental Confirmation
Recently the experiment to test quantum-like behaviour of cognitive systems
described in the previous section (see my preprint quant-ph/ 0205092,v1) was
performed by the group of Elio Conte, Center for studying of Radioactivity,
University of Bary. The experimental results obtained by this group con-
firmed the hypothesis on quantum-like probabilistic behaviour of cognitive
systems.
Description of the experiment of E. Conte:
All major aspects of cognition include perception (i.e., this is the process
of reading words or identifying objects), attention (i.e.,this is focusing on one
word at a time out of many on a page), memory (i.e., remembering the words
later), decision making (i.e., should I take A or B?), problem solving (i.e.,
figuring out what I need to do something) and language (i.e., making sense
of the words).
Owing to the complexity of the matter it is natural to start with a first
phase of experimentation based only on the most simple choice of some well
accepted tests. We start with visual perception and to this regard we arrive to
select two well consolidated tests A and B based on two different geometrical
figures and each test requiring to ask for A + or – and similarly for B + or –.
No other alternative answers resulted to be possible for A and B respectively.
A computer program was realized in order to present the figures and soon
after (disappearing the figures) requiring answers for A or for B geometrical
questions. A group of fifty subjects was selected from a larger group. Se-
lection was performed on the basis of some clinical parameters to account
for conditions of physical and mental health. Care was placed to realize a
final homogeneous group for age, cultural level, regional origin ,....., and so
on. The sex resulted rather equally distributed between males and females.
A particular care was placed to fix the times for inspection of the tests and
for answers. Automatically the computer did not take in consideration an-
swers given after a prefixed time as well as the geometrical figures appeared
only for a prefixed time: both such parameters were fixed on the basis of
neurophysiological and mental evaluations and the attempt was to explore
actually (by tests A and B and A/B) only the cognitive dynamics. Tests

191
were performed asking to a group to answer for A, asking to answer for B
and asking to answer for A having previously answered for B – A/B. All the
answers were automatically recorded and then the probabilities were calcu-
lated. Note that, using the computer, the question posed in A as well as that
one posed in B and still that one of A/B were simultaneously asked to the
groups of fifty subjects as well as all the answers were simultaneously given
within a prefixed time.
a/b a/b
Probabilities pa± were calculated as well as pa± and p++ , ..., p−− . Differences
about 3.39 percent were obtained with regard to pa± .
In the second experimentation there was considered the limit case pb− = 0.
There were obtained the following probabilities:

pa+ = 0.5714, pa− = 0.4286; pa+ = 1, pb− = 0;

a/b a/b
p++ = 0.7000, p−+ = 0.3000,
a/b a/b
p+− = 0.0000, p−− = 0.0000.
In this case there was obtained a very relevant difference of 0.5714 respect
to 0.7000.
Also if the tendency exhibited from the results seems to be unquestion-
able, the phenomenon that we are called to confirm is too important to be
liquidated by one or two experimental runs.
[Frequency Analysis of Foundations of Quantum Mechanics] Frequency
Analysis of
Foundations of Quantum Mechanics
You believe in a dice-playing God and I in perfect laws in the world of
things existing as real objects, which I try to grasp in a wildy speculative
way. A. Einstein.
Where does the Hilbert space H come from? Why does the probabil-
ity have the postulated form? Why does a physical theory which must give
real-valued results involve a complex amplitude state? Why must a quantum
particle exhibit wave behaviour (wave-particle dualism)? Must quantum me-
chanics be nonrealistic (a quantum system only has properties when they are
observed)? Is there a realistic solution of the EPR problem ? S. Gudder.
In this chapter we present frequency analysis of the contextual approach
to quantum probabilities. The consequences of this analysis coincide with the
consequences obtained in chapter 9. However, frequency analysis gives bet-
ter understanding of the balance of frequencies produced by various contexts
(preparation procedures). We also underline that physical observables con-
sidered in this chapter can, in principle, be considered as objective properties
of physical systems (as was claimed by Einstein).

192
55 Classification of Transformations of Proba-
bility
Let S be some preparation procedure (context, complex of physical con-
ditions) which produces physical systems having two properties A and B.
These properties are describing by dichotomous variables A = a1 , a2 and
B = b1 , b2 . We fix the number of preparation acts, N ≡ NS . So S always
produces ensembles S having N = |S| elements.
Let S1 and S2 be two other preparation procedures. It is assumed that
each of these preparation procedures can be applied to elements of the en-
semble S. By application of Si to S we produce a new statistical ensemble
Si , i = 1, 2 In general we need two different ensembles S to produce two
ensembles S1 and S2 .. The main feature of the ensemble Si is that B = bi
for its elements (i = 1, 2). For example, the preparation procedures Si can be
considered as filters with respect to the property B. The preparation proce-
dure Si selects elements of the ensemble S such that B = bi (i = 1, 2). Such
a filtration justifies the assumption that the number of elements in the en-
semble Si can be chosen equal to the number Ni of elements in the ensemble
S having the property B = bi (i = 1, 2). So everywhere below:

|Si | = Ni , i = 1, 2.

The crucial point of our considerations is that in general we can not select, for
example, elements with the property B = b1 without disturbing the property
A. In general the sub-ensemble

Sij = {s ∈ Si : A = aj }

of the ensemble Si does not coincide with the sub-ensemble

(0)
Sij = {s ∈ S : B = bi , A = aj }

of the original ensemble S. We set

(0)
nij = |Sij | and mij = |Sij |

(the numbers of elements in the sub-ensembles) and

Ni = |{s ∈ S : B = bi }|, nj = |{s ∈ S : A = aj }|

(the numbers of elements in S having, respectively, properties B = bi , i = 1, 2

and A = aj , j = 1, 2). We note that everywhere below the first number, i, in
the index pair ij is related to the property B and the second one, j, to the

193
property A. We shall use the frequency approach to probability (see, [143]
and [117]): the probability is defined as the limit of relative frequencies when
the number of trials N → ∞.
Remark 1.1. (Foundations of probability and physics) As we have al-
ready discussed in chapter 9 and in [117] the conventional probability theory
based on Kolmogorov’s axiomatics is not the best tool to work with ‘quantum
probabilities’. The formal use of an abstract, absolute, probability measure
is the source of many misunderstandings. In particular, the Kolmogorov
model is not the best one for operating with transitions from one context to
another. In fact, all probabilities are conditional probabilities; there is no
absolute probability. We prefer to work with frequency probabilities. Here
contexts are describing by collectives (random sequences) which are used to
find relative frequencies. However, in this book we will not pay much atten-
tion to the mathematical details of the frequency framework, see [117] for
the details. In fact, everybody who is familiar with von Mises’ frequency
probability theory can recognize that in this paper we work with von Mises’
collectives. These collectives are produced by different preparation proce-
dures (complexes of physical conditions). It may be even better to use the
term ‘collective’ instead of the term ‘ensemble’ which is used in this chapter.
However, we are little bit afraid to do this, because there is a rather strong
prejudice against von Mises’ approach.
We consider relative frequencies:
(N ) nj (N ) Ni
qj ≡ paj (N ) = , pi ≡ pbi (N ) =
N N
(for the properties A and B in the ensembles prepared by S); and relative
frequencies
a/b mij
pij (Ni ) ≡ pij (Ni ) =
Ni
(for the property A = aj in the ensemble Si prepared by Si ). We also
introduce the corresponding probabilities:
qj ≡ paj = lim paj (N );
N →∞

pi ≡ pbi = lim pbi (N );

N →∞
a/b
pij ≡ pij = lim pij (N ).
Ni →∞

Because in general the nij are not equal to mij (even asymptotically, when
N → ∞) we do not have the conventional formula of total probability. In
general
qj = pS (A = aj ) 6= p1 p1j + p2 p2j .

194
We want to investigate the various forms probabilities qj can take, depending
on perturbations induced by context transitions. In the general case we have:
nj n1j n2j
qj (N ) = = +
N N N
m1j m2j (N ) N1 m1j N2 m2j (N )
= + + δj = . + . + δj
N N N N1 N N2
(N )
= p1 (N )p1j (N1 ) + p2 (N )p2j (N2 ) + δj ,
where the perturbation term (which appears as a result of the transition from
the context S to new contexts S1 and S2 ) has the form:

(N ) 1
δj ≡ δ(S → S1 , S2 ; j; N ) = [(m1j − n1j ) + (m2j − n2j )].
N
We remark that there always exists the limit

δj = lim δj (N ) = qj − (p1 p1j + p2 p2j ).

N →∞

Thus in general we have

qj = p1 p1j + p2 p2j + δj ,

where
1
δj = lim [(m1j − n1j ) + (m2j − n2j )].
N →∞ N

It is useful to carry out normalization by setting

√
δj = 2 p1 p1j p2 p2j λj , j = 1, 2.

The trivial (but important) remark is that there are three possibilities (com-
pare to chapter 9):
(T) |λj | ≤ 1;
(H) |λj | > 1.
(HT) |λ1 | ≤ 1 and |λ2 | > 1 or |λ1 | > 1 and |λ2 | ≤ 1.
In the case (T) we can always represent the coefficient as λj = cos θj
and in the case (H) as λj = ± cosh θj , j = 1, 2, and in the case (HT) as
λ1 = cos θ1 and λ2 = ± cosh θ2 or vice versa. Probabilistic behaviours of
the types (T), (H), and (HT) will be called trigonometric, hyperbolic and
hyper-trigonometric behaviours, respectively.
We have studied the general case. There are three preparation procedures
S, S1 , and S2 such which S1 and S2 are selections with respect to values B =

195
b1 and B = b2 (for elements of the ensemble S prepared by the preparation
procedure S). As always, we need two copies of the ensemble S to perform
two selection procedures, S1 and S2 . The general probabilistic transformation
induced by transitions S → Sj , j = 1, 2, has the form:
q
a b a/b b a/b a/b a/b
pj = p1 p1j + p2 p2j ± 2 pb1 p1j pb2 p2j λj , (125)

where λj = cos θj or λj = cosh θj , or λ1 = cos θ1 and λ2 = ± cosh θ2 or vice

versa. Here the coefficient λj gives the normalized statistical measure of the
perturbations of A induced by the transitions S → S1 , S2 :
(N )
λj = lim λj ,
N →∞

where
(N ) 1
λj = √ [n1j − m1j ) + (n2j − m2j )]. (126)
2 m1j m2j
If these perturbations are relatively small, namely, |λj | ≤ 1, j = 1, 2, then
we observe T-behaviour; in particular, classical and quantum behaviours. If
these perturbations are relatively large, namely, |λj | > 1, j = 1, 2, then we
observe H-behaviour. In fact, we can continuously transform T-behaviour
into H-behaviour, since λj , |λj | = 1, has both T- and H-representations:
λj = ± cos 0 = ± cosh 0. If one of these perturbations, for instance λ1 , is
relatively small, namely, |λ1 | ≤ 1, and another, λ2 , is relatively large, namely,
|λ2 | > 1, then we observe HT-behaviour.
Finally, we show that the coefficients λ1 and λ2 are connected by a ‘con-
dition of orthogonality’. In the quantum formalism this is the real condition
of orthogonality in the complex Hilbert space. We note that the matrix of
probabilities P = (pij ) is always a stochastic matrix (see also chapter 9):

p11 + p12 = 1 and p21 + p22 = 1. (127)

Thus we have
a/b a/b a/b a/b
1 = pa1 + pa2 = pb1 p11 + pb2 p21 + pb1 p12 + pb2 p22
q q
b a/b b a/b a/b a/b
+2 p1 p11 p2 p22 λ1 + 2 pb1 p12 pb2 p22 λ2 .
To simplify considerations we assume everywhere which all probabilities are
strictly positive. This implies:
q q
a/b a/b a/b a/b
p11 p21 λ1 + p12 p22 λ2 = 0 . (128)

196
We set v
u a/b a/b
up p
K = t 12 22
a/b a/b
.
p11 p21
We obtain
λ1 = −Kλ2 .
We observe that the probabilities pbj are not involved in the orthogonality
condition (128). In particular, in the T -case we always have:

cos θ1 = −K cos θ2 , (129)

in the H-case we have:

cosh θ1 = K cosh θ2 (130)
(here λ1 = ± cosh θ1 and λ2 = ∓ cosh θ2 ).
In the HT-case we have:

cos θ1 = ±K cosh θ2 or cosh θ1 = ±K cos θ2 . (131)

Finally, we remark that all above the considerations can be easily gener-
alized to non-dichotomous variables: A = a1 , ..., aM and B = b1 , ..., bM .
In this case the probability pai can always be represented as
M
X Xq
a/b a/b a/b (i)
pai = pbj pji +2 pbk pbl pki pli λkl , (132)
j=1 khl

where the coefficients are given by

(i)
(i) δkl
λkl = q
a/b a/b
2 pbk pbl pki pli

and
(i) 1 a/b a/b
δkl = [pbk (pai − pki ) + pbl (pai − pli )] .
M −1
(i)
The coefficients {λkl } are normalized statistical deviations which arise as
a consequence of the transition from the context determined by S to the
contexts Sj .
To simplify analysis we shall consider only dichotomous variables in the
following sections.

197
56 Classical, Quantum, and Non-Classical-Quantum
Physics
In this section we consider probabilistic transformations for preparation pro-
cedures which produce relatively small statistical deviations:

|λj | ≤ 1, j = 1, 2.

2.1. Classical probabilistic behaviour.

Suppose which we can construct statistically ‘perfect’ preparation proce-
dures S1 , S2 : selections of elements of the ensemble S with respect to values
B = b1 and B = b2 produce statistically negligible changes of A. We set

∆ij (N ) = nij − mij .

Here, as always, nij is the number of elements of the ensemble S for which
B = bi and A = aj and mij is the number of elements of the ensemble Si for
which A = aj . The classical probabilistic behaviour is characterized by the
condition:
∆ij (N )
lim = 0 for all i, j.
N →∞ N
Here both λj = 0, and we have the conventional formula of total probability.
2.2. Quantum probabilistic behaviour.
Let us consider preparations which induce symmetric statistical devia-
tions:
|λ1 | = |λ2 | . (133)
Thus the coefficient K is equal to 1. So p12 p22 = p11 p21 . In the two-
dimensional case this condition is equivalent to the well known condition
of double stochasticity (see chapter 9):

p11 + p21 = 1, p12 + p22 = 1. (134)

These are ‘conservation laws’ for the A in the process of splitting of the en-
semble S into ensembles S1 and S2 . We also remark that (129) implies which
cos θ1 = − cos θ2 . So θ2 = θ1 + π (mod 2π). Thus we have the probabilistic
transformations:
√
q1 (≡ pa1 ) = p1 p11 + p2 p21 + 2 p1 p11 p2 p21 cos θ , (135)
√
q2 (≡ pa2 ) = p1 p12 + p2 p22 − 2 p1 p12 p2 p22 cos θ . (136)

198
This is the well known quantum probabilistic transformation. We now find
complex representations of these probabilities which would linearize trans-
formations (135), (136). We use the well known formula:
√ √ √
A + B ± 2 AB cos θ = | A ± Beiθ |2 . (137)

Thus
√ √ √ √ √ √ √ √
q1 = | p1 p11 + p2 p21 eiθ1 |2 ; q2 = | p1 p21 + p2 p22 eiθ2 |2 .

(in quantum case θ1 = θ2 +π). These formulas can be also derived by C-linear
space computations.
We represent the preparation procedure S by a vector ϕ in the two-
dimensional complex Hilbert space:
√ √
ϕ = p1 ϕ1 + p2 eiθ ϕ2 ,

where {ϕ1 , ϕ2 } is an orthonormal basis corresponding to the physical observ-

able B (the condition p1 + p2 = 1 implies which ||ϕ||2 = 1). Let ψ1 , ψ2 be an
orthonormal basis corresponding to the physical observable A. We have:
√ √ √ √
ϕ1 = p11 ψ1 + eiγ1 p12 ψ2 , ϕ2 = p21 ψ1 + eiγ2 p22 ψ2 .

We remark that the orthogonality of ϕ1 and ϕ2 is, in fact, equivalent to

the condition of double stochasticity for P = (pij ) and the relation γ2 =
γ1 + π (mod 2π). By expanding ϕ with respect to the basis {ψ1 , ψ1 } we
obtain
ϕ = α1 ψ1 + α2 ψ2 ,
where
√ √ √ √ √ √
α1 = p1 p11 + eiθ p2 p21 , α2 = eiγ1 p1 p12 + ei(γ2 +θ) p2 p22 . (138)

By using the relation γ2 = γ1 + π we reproduce the quantum probabilistic

rule (135), (136).
We note that our considerations have demonstrated that the main distin-
guishing feature of quantum formalism is not the presence of the factor cos θ
in the ‘quantum transformation of probabilities’, but the double stochasticity
a/b
of the matrix P = (pij ) of transition probabilities and the relation

γ2 = γ1 + π (139)

between phases in the expansions of ϕ1 and ϕ2 with respect to the basis

{ψ1 , ψ2 }.

199
 The ‘double stochasticity conservation laws’ (134) and the ‘phase con-
servation law’ (139) imply the unitarity of the transformation U connecting
{ϕ1 , ϕ2 } and {ψ1 , ψ2 }. In fact, this is the root of the superposition principle
(see the next subsection for the details).
Finally, we remark that there is a crucial difference between classical
physical behaviour (λ1 = λ2 = 0) and quantum decoherence (λ1 = λ2 = 0).
In the first case the coefficients λj = 0 because statistical deviations are
negligibly small. In the second case the coefficients λj = 0 because statistical
deviations compensate each other (j = 1, 2) :
∆1j ∆2j
≈− , N → ∞.
N N
2.3. Neither classical nor quantum trigonometric behaviour.
Here the matrix P = (pij ) of transition probabilities need not be double
stochastic. We can find the probability distribution qj = paj = pS (A =
aj ), j = 1, 2, by using the following transformation of probabilities:
√
qj = p1 p1j + p2 p2j + 2 p1 p1j p2 p2j cos θj , (140)
q
where cos θ1 = −K cos θ2 , K = pp12 p22
11 p21
. In general such a probabilistic trans-
formation (‘interference’ between the preparation procedures S1 and S2 ) can
not be described by the standard (Dirac – von Neumann) quantum formal-
ism.
Example 2.1. Let p1 = p2 = 12 (symmetric distribution of B in the
ensemble S; for example, the two slit experiment with the location of slits
symmetric with respect to the source of particles) and let p11 = p12 = 21
(symmetric distribution of A in the ensemble S1 ) and p21 = 31 , p22 = 23
(asymmetric distribution of A in the ensemble S2 ). Thus the matrix P is not
double stochastic.
The law of conservation of the A is violated in the process of the transition
S → (S1 , S2√). The measure of this violation is given by the coefficient K.
Here K = √ 2. The phases θ1 and θ2 must be chosen in such a way that
cos θ1 = − 2 cos θ2 . For example, we can consider preparations such that
θ1 = 3π/4 and θ2 = π/3. In this case we have
5 cos 3π 7 cos π
pa1 = + √ 4 ; pa2 = + √3 .
12 6 12 3
This probabilistic transformation can not be obtained in standard ‘quantum
linear calculus’In this (Dirac – von Neumann) calculus physical observables
should be represented by self-adjoint operators.. We shall see that it can be
obtained by non-unitary generalization of ‘quantum linear calculus’.

200
57 Hyperbolic Probabilistic Behaviour
In this section we consider examples of H-behaviour and HT-behaviour. We
remark that H-behaviour can be exhibited by preparations having double
stochastic transition matrices.
Example 3.1. Let p1 = α and p2 = 1 − α (0 < α < 1) and let
pij = 1/2, i, j = 1, 2. Here K = 1 (the transition matrix is double stochastic)
and, hence, cosh θ2 = cosh θ1 . We have
1 p
q1 = + α(1 − α) cosh θ,
2
1 p
q2 = − α(1 − α) cosh θ,
2
In the opposite to the T-case the phase θ cannot take arbitrary values. There
is a relation between θ and α which provides that q1 , q2 have the meaning of
probabilities. We set
1
e(α) = p .
2 α(1 − α)
We remark that e(α) ≥ 1 for all 0 < α < 1. The hyperbolic phase θ
can be chosen as θ ∈ [0, θmax ], where θmax = arccosh e(α). For example,
let α = 14 (1 − α = 34 ). Thus e(x) = √23 . Here we can observe hyperbolic
interference for angles 0 ≤ θ ≤ arccosh √23 . We remark that if p1 = p2 = 12
then e(α) = 1 and the hyperbolic interference coincides with the ordinary
interference cos 0 = cosh 0 = 1. In general the symmetric distribution p1 =
p2 = 12 can produce nontrivial hyperbolic interference. We have for general
double stochastic matrix P :
1 √ 1 √ 1 p
qj = (p1j + p2j ) + p1j p2j λj = + p1j p2j λj = + α(1 − α)λj ,
2 2 2
where we set α = p11 = p22 and 1 − α = p12 = p21 . If θ ∈ [0, θmax ], θmax =
arccosh e(α), then λj = ± cosh θ, θ 6= 0.
We remark that the total symmetry (in S as well as S1 , S2 ), namely
p1 = p2 = pij = 21 , produces the trivial H-interference (which coincides with
the T-interference). So hyperbolic interference might be observed only for
preparation procedures with asymmetric probability distributions for con-
texts.
Remark 3.1. (Negative probabilities) If we do not pay attention to
the range of the H-phase parameter θ we can obtain negative probabilities
and probabilities which are larger than one. It must be noted that such

201
‘probabilities’ appear with an intriguing regularity in various extensions of
the quantum formalism (Wigner [194], Dirac [63], Feynman [74], see also
[117] for the details). It may be that ‘quantum negative probabilities’ have
the same origin as ‘negative H-probabilities,’ namely, the use of nonphysical
values of some parameters, see [117] for the details.
Of course, our considerations induce the following natural question: ‘Is
it possible to construct a linear space representation for the H-probabilistic
transformations?’ We shall study this question in section 5.
Finally, we consider an example of mixed HT-behaviour.
Example 3.2. Let p1 = p2 = 12 and let p11 = 45 , p12 = 51 , p21 = 45 , p22 =
1
5
. We have K = 14 ; so λ2 = −4λ1 .
We have q1 = 45 (1 + λ1 ), q2 = 15 (1 − 4λ1 ). If −1 ≤ λ1 ≤ 14 , then q1 and
q2 have the meaning of probabilities. For example, let λ1 = − 21 and λ2 = 2.
Then q1 = 25 , q2 = 53 . Thus
4 4 2 1 1 √
q1 = + cos π, q2 = + cosh(ln(2 + 3)).
5 5 3 5 5
We remark that mixed HT-behaviour can not be produced on the basis of a
double stochastic matrix P = (pij ).
Finally, we note that the H-phase has a symmetry, θ → −θ, which is an
analogue of the symmetry θ → θ + 2π for the T-phase. If λ = cosh θ, then θ
can be chosen as
√ √
θ = ln(λ + λ2 − 1) or θ = ln(λ − λ2 − 1) .

58 Linear Space Representation of the Trigono-

metric Rule
We shall study the possibility of representing the general probabilistic trans-
formation (140) as a linear transformation in a complex linear space. Because
in general the transition probability matrix P = (pij ) is not double stochastic
we can not expect that it would be possible to work with orthonormal bases
in a complex Hilbert space. It seems which the inner product structure is
not useful in the general case.
Let E be a two-dimensional linear space over the field of complex num-
bers C. The choice of C as the basic number field has a trivial explanation.
Formula (137) gives the possibility of representing the T-probabilistic trans-
formation in form (138) which is reduced to the transition from one basis
to another. It is impossible to linearize the quantum probabilistic transfor-
mation by using real numbers, but it is possible to do so by using complex

202
numbers. These arguments were already evident in our analysis of quantum
theory. We now observe that they can be used in a more general situation.
Vectors of E are said to be quantum states. At the moment there is
no Hilbert structure on E. There is nothing similar to the standard nor-
malization condition for quantum states. We represent the ensemble S (the
preparation procedure S) by a vector ϕ in E; the ensembles S1 and S2 (the
preparation procedures S1 and S2 ) by vectors ϕ1 and ϕ2 .
It is supposed that the preparation procedures S1 and S2 determine some
dichotomous physical variable, B = b1 , b2 . In the linear space calculus this
assumption has the following counterpart: vectors {ϕ1 , ϕ2 } are linearly inde-
pendent in E.
The splitting of S into S1 and S2 (owed to the preparation procedures
S1 and S2 ) is represented as expanding the vector ϕ with respect to a basis
{ϕ1 , ϕ2 } in E. We can always expand the vector φ with respect to the basis:

ϕ = β1 ϕ1 + β2 ϕ2 ,

where β1 and β2 ∈ C. As in the ordinary quantum formalism the probabilities

pbi = PS (B = bi ) are represented as pbi = |βi |2 (a generalization of Born’s
postulate). So there is a constraint for vectors ϕ and ϕ1 , ϕ2 :

|β1 |2 + |β2 |2 = 1 . (141)

In such a case the quantum state ϕ is said to be B-decomposable.

We now consider the measurement of A for ensembles Si (prepared by Si ).
We consider a measurement such that a second measurement of A, performed
immediately after the first, will yield the same value of the observable. Thus
such an A-measurement can be interpreted as a preparation procedure.
To be more precise, we consider two preparation procedures S1a and S2a
corresponding to selections of physical systems on the basis of the values
A = a1 and A = a2 . The B-preparation procedures S1 and S2 we now denote
by the symbols S1b and S2b , respectively.
The Sja selects physical systems such that A = aj , j = 1, 2. We remark
that in general these selections may change the probability distribution of
B. By applying Sja to the ensemble Sib ≡ Si (which was produced by the
application of Sib to an ensemble S produced by S) we obtain an ensemble
Sijba , i, j = 1, 2. In the same way we split the ensemble S (with the aid
of S1a and S2a ) into ensembles Sja , j = 1, 2. Ensembles Sja , j = 1, 2, are
represented by vectors ψj in the E. We assume that they also form a basis in
E (this is a consequence of the preparation procedures S1a and S2a determining
the dichotomous physical variable A). Thus splitting S → (S1a , S2a ) can be

203
represented by the expansion
ϕ = α1 ψ1 + α2 ψ2 ,
where αj ∈ C. Here the probabilities paj = PS (A = aj ) = |αj |2 , so
|α1 |2 + |α2 |2 = 1. (142)
Thus ϕ is A-decomposable.
In the general case we have to represent the ensembles Sijba , i, j = 1, 2,
by four different vectors ψij . In general we cannot assume that these vectors
belong to the same two-dimensional space E. The study of this general situ-
ation is too complicated. We restrict ourselves to the special case (which is
the most interesting for applications). Let ψ11 = ψ1 , and ψ21 = ψ1 , ψ21 = ψ2
and ψ22 = ψ2 . It was assumed that ψ1 and ψ2 are independent vectors.
We would like to predict the probabilities paj on the basis of the transition
from the basis {ϕ1 , ϕ2 } to the basis {ψ1 , ψ2 }. Let U = (uij ) be the transition
matrix (at the moment the only restriction on U is its invertibility). It is
supposed which each vector ϕi is A-decomposable In general there is no com-
position (rather it would be better to say decomposition) transitivity. For
example, it may be that the state ϕ is B-decomposable and each state ϕi is
A-decomposable, but ϕ is not A-decomposable. We suppose decomposabil-
ity of all states under consideration for physical reasons: the possibility of
performing A and B measurements for elements of S. The violation of com-
position transitivity corresponds to the following situation: we can perform
B-measurements on S and A-measurements on Sib , but we can not perform
A-measurement on S.. Thus
|ui1 |2 + |ui2 |2 = 1, i = 1, 2. (143)
We have
α1 = β1 u11 + β2 u21 , α2 = β1 u12 + β2 u22 . (144)
Coefficients βj , uij are not independent. They satisfy the constraint (142).
Simple computations give us
β1 β̄2 (u11 ū21 + u12 ū22 ) + β̄1 β2 (ū11 u21 + ū12 u22 ) = 0. (145)
One of the solutions of this equation is given by
u11 ū21 + u12 ū22 = 0. (146)
This is the condition of unitarity of the transition matrix U = (uij ). This
solution gives the ordinary quantum formalism. In this formalism it is useful
to introduce the inner product:
hz, wi = z1 w̄1 + z2 w̄2

204
and rewrite the above equation as the condition of orthogonality of vectors
ϕ1 and ϕ2 : hϕ1 , ϕ2 i = 0. However, equation (145) has other solutions which
are not related to the standard quantum formalism. These solutions give the
complex linear space representation for the trigonometric probabilistic rule
in the nonclassical/quantum case. We set:
√ √
βi = pi eiξi , uij = pij eiγij ,

where p1 + p2 = 1, p11 + p12 = 1, p21 + p22 = 1 and ξ1 , γij are arbitrary

phases. Thus the transition from one basis to another has the form:
√ √ √ √
ϕ1 = p11 eγ11 ψ1 + p12 eγ12 ψ2 , ϕ2 = p21 eγ21 ψ1 + p22 eγ22 ψ2 . (147)

In these notations equation (145) has the form:

√ √
cos(η + γ1 ) p11 p21 + cos(η + γ2 ) p12 p22 = 0, (148)

where η = ξ1 − ξ2 , γ1 = γ11 − γ21 , γ2 = γ12 − γ22 .

We set θ1 = η+γ1 and θ2 = η+γ2 . Equation (148) coincides with equation
(128) in the T-case. Thus all possible probabilistic T-transformations can be
represented in the complex linear space. A rather surprising fact is which
equation (148) has a new (non-quantum solution) even for a double stochastic
matrix of transition probabilities.
Let P be a double stochastic matrix. Equation (148) has the form:

cos(η + γ1 ) + cos(η + γ2 ) = 0 .

Thus
(2η + γ1 + γ2 ) (γ1 − γ2 )
cos = 0 or cos =0.
2 2
There is a crucial difference between these equations. The first equation
‘remembers’ the state ϕ; splitting ϕ into {ϕ1 , ϕ2 } (or S into S1 and S2 ).
This memory is given by the phase shift η. The second equation does not
contain any memory term. In fact, this is the standard quantum mechanical
equation: γ1 − γ2 = π (mod 2π).
Thus we obtain a new (non-quantum) solution even for a double stochastic
matrix P = (pij ) :
2η + γ1 + γ2 = π (mod 2π).
In this case the transformation (147) also reproduces the quantum proba-
√
bilistic rule (135), (136): qj = p1 p1j + p2 p2j ± 2 p1 p1j p2 p2j cos θ. However,
(147) is not unitary:

u11 β̄21 + u12 β̄22 = 1 − e−2iη 6= 0, η 6= 0.

205
59 Linear Space Representation of the Hyper-
bolic Rule
We want to find a kind of linear space calculus for the H-probabilistic trans-
formation. It seems that it would be impossible to do this in a C-linear
space. We propose to use a hyperbolic algebra G — the two-dimensional
Clifford algebra, see, e.g., [114]. This is a two-dimensional real algebra with
basis e0 = 1 and e1 = j, where

j 2 = 1.

Elements of G have the form z = x + jy, x, y ∈ R. We have z1 + z2 =

(x1 + x2 ) + j(y1 + y2 ) and z1 z2 = (x1 x2 + y1 y2 ) + j(x1 y2 + x2 y1 ). This algebra
is commutative. We introduce an involution in G by setting z̄ = x − jy. We
set
|z|2 = z z̄ = x2 − y 2 .
p
We remark that |z| = x2 − y 2 is not well defined for an arbitrary z ∈ G.
We set G+ = {z ∈ G : |z|2 ≥ 0}. We remark that G+ is the multiplicative
semigroup (z1 , z2 ∈ G+ → z = z1 z2 ∈ G+ ). It is a consequence of the equality
|z1 z2 |2 = |z1 |2 |z2 |2 .
Thus for z1 , z2 ∈ G+ we have |z1 z2 | = |z1 ||z2 |. We introduce

ejθ = cosh θ + j sinh θ, θ ∈ R.

We remark that

ejθ1 ejθ2 = ej(θ1 +θ2 ) , ejθ = e−jθ , |ejθ |2 = cosh2 θ − sinh2 θ = 1.

Hence z = ±ejθ always belongs to G+ . We also have

ejθ + e−jθ ejθ − e−jθ

cosh θ = , sinh θ = .
2 2j

We set G∗+ = {z ∈ G+ : |z|2 > 0}. Let z ∈ G∗+ . We have

x y xsignx ysignx
z = |z|( + j ) = sign x |z| ( +j ).
|z| |z| |z| |z|
2 2
y
x
During |z| 2 − |z|2 = 1, we can represent x sign x = cosh θ and y sign x = sinh θ,

where the phase θ is unequally defined. We can represent each z ∈ G∗+ as

z = sign x |z| ejθ .

206
By using this representation we can easily prove that G∗+ is the multiplicative
group. Here z1 = signx
|z|
e−jθ . The unit circle in G is defined as S1 = {z ∈ G :
|z|2 = 1} = {z = ±ejθ , θ ∈ (−∞, +∞)}. It is a multiplicative subgroup of
G∗+ .
Hyperbolic Hilbert space is G-linear space (module, see [114]) E with a
G-scalar product, a map h·, ·i : E × E → G, that is
1) linear with respect to the first argument:

haz + bw, ui = ahz, ui + bhw, ui, a, b ∈ G, z, w, u ∈ E;

2) symmetric: hz, ui = hu, zi;
3) non-degenerate: hz, ui = 0 for all u ∈ E iff z = 0.
We note that 1) and 2) imply which

hu, az + bwi = āhu, zi + b̄hu, wi.

Remark 5.1. If we consider E as just an R-linear space then h·, ·i is a

(rather special) bilinear form which is not positive definite. In particular, in
the two-dimensional case we have the signature (+, −, +, −).
We represent the H-probabilistic transformation in the two-dimensional
G-linear space (module) E. From the beginning we do not consider any
G-Hilbert structure on E. Such a structure will appear automatically in
the representation of one particular class of H-probabilistic transformations,
the H-quantum formalism. In the same way as in the previous section we
introduce quantum states ϕ, {ϕ1 , ϕ2 }, {ψ1 , ψ2 } corresponding to preparation
procedures (statistical ensembles). By definition a quantum state is a vector
belonging to a G-linear space (no normalization!).
It is supposed that {ϕ1 , ϕ2 } and {ψ1 , ψ2 } are bases in the G-linear space
E. It is also supposed that the state ϕ is B and A-decomposable and the
states ϕi are A-decomposable. Thus:

ϕ = β1 ϕ1 + β2 ϕ2 , |β1 |2 + |β2 |2 = 1, |βj |2 ≥ 0,

and
ϕ1 = u11 ψ1 + u12 ψ2 , ϕ2 = u21 ψ1 + u22 ψ2 ,
where vectors of coefficients u(1) = (u11 , u12 ) and u(2) = (u21 , u22 ) are such
that
|u11 |2 + |u12 |2 = 1, |u21 |2 + |u22 |2 = 1 and |uij |2 ≥ 0.
Thus
ϕ = α1 ψ1 + α2 ψ2 ,

207
where the coefficients α1 , α2 are given by (144). There is no formal difference
between linear space transformations over C and G. However, the assump-
tion which the state ϕ is A-decomposable implies which the G-linear space
calculations have a physical meaning iff the vector α = (α1 , α2 ) is such that

|α1 |2 = |β1 u11 + β2 u21 |2 ≥ 0, |α2 |2 = |β1 u12 + β2 u22 |2 ≥ 0, (149)

and
|α1 |2 + |α2 |2 = 1. (150)
The latter equation coincides with equation (142) (with the only difference
which all numbers belong to G instead of to C).
As we have already discussed in the T-case, in general there is no compo-
sition (in fact, decomposition) transitivity. In general the B-decomposability
of ϕ and the A-decomposability of ϕi need not imply that ϕ is also A-
decomposable. Our assumptions on composition transitivity are based on
the physical context of our considerations.
As in the T-case, (145) has the solution given by equation (146) (the
only difference is that now all coefficients belong to the hyperbolic algebra).
This is the condition of orthogonality of the vectors ϕ1 and ϕ2 with respect
to the G-linear product:hz, wi = z1 w̄1 + z2 w̄2 . So the matrix U = (uij ) is a
G-unitary matrix, namely

hu(i) , u(j) i = δij . (151)

We now study the general case. Here the U need not be a unitary matrix.
We consider only vectors with coefficients belonging to G∗+ . We set βi =
√ √
± pi ejξi , uij = ± pij ejγij , i, j, = 1, 2. Condition (150) is equivalent to the
condition:
√ √
p12 p22 cosh θ2 + σ p11 p21 cosh θ2 = 0,
where σ = uij signuij . This equation has a solution, namely, the phases θ1
and θ2 , iff
σ = −1. (152)
Thus the transition matrix U=(uij ) must always satisfy (152).
Let us turn back to the case in which U is a G-unitary matrix. We shall
call such a model hyperbolic quantum formalism. The orthogonality relation
implies:
0 = hu(1) , u(2) i =
√ √
signu11 signu21 p11 p21 ej(γ11 −γ21 ) + signu12 signu22 p12 p22 ej(γ12 −γ22 ) ;
or
1 + σKej(γ1 −γ2 ) = 0,

208
where
√ √
K= p12 p22 / p11 p21 and γ1 = γ12 − γ22 , γ2 = γ11 − γ21 .

Thus sinh(γ1 − γ2 ) = 0, and hence we obtain:

γ1 = γ2 (153)

(we recall which in the standard quantum formalism we have γ1 = γ2 +

π (mod 2π)). We also have

1 + σK cosh(γ1 − γ2 ) = 0.

Thus σ = −1 and K = 1. So the sign-condition (152) is always satisfied

for a unitary matrix U=(uij ). The equality K = 1 is equivalent to double
stochasticity of the matrix of transition probabilities P = (pij = |uij |2 ).
Therefore the matrix U = (uij ) is a G-unitary matrix iff the corresponding
matrix of probabilities P = (pij ) is a double stochastic matrix, σ = −1, and
the hyperbolic phases satisfy (153).
The H-quantum formalism (special calculus in a G-linear space) repre-
sents probabilistic transformations
√
q1 = p1 p11 + p2 p21 ± 2 p1 p2 p11 p21 cosh θ,
√
q2 = p1 p12 + p2 p22 ∓ 2 p1 p2 p12 p22 cosh θ,
where θ = γ11 − γ21 = γ12 − γ22 .
The situation is similar to the ordinary quantum formalism. However,
there is the important difference between these formalisms. In the T-quantum
formalism the condition of C-unitarity of U = (uij ) was also sufficient to
obtain a physically meaningful transformation of probabilities: all possi-
ble phases θ give a meaningful probabilistic transformation for the fixed
C-unitary matrix U = (uij ). It is not so in the H-quantum formalism. The
G-unitary of U = (uij ) is not sufficient to obtain physically meaningful prob-
abilities for all H-phases θ. Besides condition (150) we also have condition
(149) which provides non-negativity of probabilities qj = paj = |αj |2 .
We set t = p11 = p22 (so p12 = p21 = 1 − t), 0 < t < 1 (we recall that
P = (pij ) is a double stochastic matrix). We also set p1 = s, so p2 = 1−s, 0 <
s < 1. Let us consider the case in that sign u11 sign u21 = −1. Hence sign
u12 sign u22 = 1. Here
p
q1 = st + (1 − s)(1 − t) − 2 s(1 − s)t(1 − t) cosh θ,
p
q2 = s(1 − t) + (1 − s)t + 2 s(1 − s)t(1 − t) cosh θ.

209
Thus
st + (1 − s)(1 − t)
cosh θ ≤ p = e(s, t).
2 s(1 − s)t(1 − t)
Thus physical H-behaviour is possible only for probabilities s, t such that
e(s, t) ≥ 1 (in the case of the equality H and T-behaviours coincide).
We note that there is no analogue of the superposition principle in the
H-quantum formalism. G-unitary transformations preserve normalization
condition (150), but they do not preserve positivity conditions (149).
We now turn back to the general case in which the P need not be double
stochastic. We consider again equation (150) which is equivalent to (145)
(with coefficients belonging to the hyperbolic algebra). We have already
studied the special class of solutions of equation (145) given by equation
(146). These solutions are given by G-unitary matrixes. We now consider
the general equation:

σK cosh(η + γ2 ) + cosh(η + γ1 ) = 0. (154)

As σ = −1 we finally obtain the equation

K cosh θ2 = − cosh θ1

(compare with (129)).The presence of the H-phase η = ξ1 − ξ2 plays the role

of memory in the preparation procedure S (which produced an ensemble S
represented by the state ϕ). We remark that equation (154) has the following
two solutions for K = 1 (double stochastic matrix):

cosh(η + γ2 ) = cosh(η + γ1 ) → η + γ2 = η + γ1 orη + γ2 = −η − γ1 .

In the first case we have the H-quantum solution,γ1 = γ2 , and in the second
case we have a new solution 2η + γ2 + γ1 = 0, which corresponds to the
non-unitary transition matrix U.

60 Conclusions
Our frequency analysis of probabilities related to transitions from one experi-
mental arrangement (context, complex of physical conditions) to another has
the following consequences:
A. Quantum rule
The quantum rule for interference of probabilistic alternatives can be
obtained in the purely contextual approach; in particular, without appealing
to wave arguments.

210
 B. Classical and quantum worlds
Both the quantum and classical probabilities can be interpreted as fre-
quency probabilities. Specific quantum behaviour of probabilities in exper-
iments with quantum particles is related to the specific relation between
elementary particles and experimental arrangement. There in the quantum
world transition from one context to another produces statistical perturba-
tions which change the classical formula of total probability (by the additive
interference term). In the classical world such perturbations are negligibly
small statistically.
C. Trigonometric and hyperbolic transformations
Transformations of probabilities corresponding to context transitions can
be classified according to the magnitudes of statistical perturbations: trigono-
metric, hyperbolic, hyper-trigonometric. In particular, contextual modifica-
tions of the classical formula of total probability are not reduced to the quan-
tum rule for interference of probabilistic alternatives. There exists trigono-
metric interference of probabilistic alternatives which is neither classical nor
quantum. There also exists hyperbolic interference. Trigonometric transfor-
mations correspond to context transitions inducing relatively small statistical
perturbations; hyperbolic - relatively large.
D. Double stochasticity
The main distinguishing feature of quantum probabilistic transformations
is not the appearance of the cos θ- interference term, but the double stochas-
ticity of the matrix of transition probabilities — the law of statistical balance.
E. Complex amplitudes
Starting with the trigonometric transformation of probabilities, we obtain
(with the aid of the cosine-theorem) the complex amplitude representation of
contextual probabilities. This gives the possibility of constructing a complex
linear space representation of contextual probabilistic calculus. In general, we
can not represent a trigonometric probabilistic transformation in a complex
Hilbert space. It is possible only for double stochastic matrices of transi-
tion probabilities (which corresponds to unitary transformations of a Hilbert
space).
F. No superposition
One of the special features of the general C-linear representation of con-
textual probabilities is the violation of the superposition principle. It seems
that this fundamental principle is a consequence of double stochasticity.
G. Hyperbolic Hilbert space

211
 Hyperbolic probabilistic transformations can be represented as linear trans-
formations in modules over the system (commutative algebra) G of hyper-
bolic numbers (the two-dimensional Clifford algebra). One of special features
of the general G-linear representation of contextual probabilities is the vio-
lation of the superposition principle. In this case even double stochasticity
of the matrix of transition probability (G-unitarity of the corresponding G-
linear transformation) does not imply the superposition principle.
H. Super-quantum scale
We obtained hyperbolic interference for complexes of physical conditions
which produce statistical perturbations of relatively large magnitude. Here
physical systems are supersensitive to perturbations produced by measure-
ment devices. Such a behaviour would be natural for physical systems having
essentially higher sensitivity (to perturbations induced by our macroscopic
devices) than ordinary quantum systems (elementary particles). We can
speak about the ‘super-quantum’ scale of sizes and energies. In particular, on
this scale there must exist a new hyperbolic Planck constant. Formally such
a parameter appears in the hyperbolic analogue of the Schrödinger equation
in hyperbolic quantum mechanics. Of course, in accordance with our general
ideology, hyperbolic interference also can be produced in some experiments
with macro systems.
I. Numerical simulation
In principle, probabilistic behaviour (trigonometric as well as hyperbolic)
that is neither classical nor quantum can be simulated numerically.
J. Superselections
Superselection rules are closely related to the superposition principle.
With a superselection rule unitarity (double stochasticity of the matrix of
transition probabilities) or linear combinations do not imply coherent su-
perposition. Superselection rules are also important since they are relevant
to macroscopic quantum systems. I think that superselection rules give re-
strictions on the physical realization of some preparation procedures, namely
filtration (selection) procedures which give a possibility of transforming an
ensemble of physical systems prepared under one fixed complex of conditions
S into an ensemble of physical systems prepared under some special complex
of physical conditions S 0 . There exist complexes S and S 0 such that it is
impossible to create the corresponding transformation procedure. However,
I think (and it may be that I am wrong) that superselection rules can not be
analysed in a general probabilistic framework. Each rule is closely connected
to some fixed class of physical systems under consideration. If we represent

212
in the same linear (in particular, Hilbert) space contextual probabilities for
distinct classes of physical systems, then we shall obtain distinct classes of
contexts which can not be transformed into each other.
K. Wave-particle dualism
It is well known that the wave-particle dualism was proposed for solving
the contradiction between the classical and quantum rules for addition of
probabilities of alternatives. The violation of the classical rule was observed
in the well known two slit experiment (by comparing results of measurements
for three different complexes of physical conditions: both slits are open, only
one of slits is open). By using the frequency probabilistic framework we
obtained ‘quantum interference’ without appealing to wave arguments. Of
course, as we have mentioned in this chapter, we can introduce complex
amplitudes of probabilities. However, this wave description is just a math-
ematical description, complex linearization of nonlinear transformations of
probabilities. So, in principle, we can use purely corpuscular terminology.
L. Quantum-like behaviour for macro systems
Our derivation of quantum interference has been purely mathematical.
The only physical constraint is that context transitions induce statistical
perturbations of relatively small magnitude. In principle, such perturbations
can be produced in experiments with macroscopic systems. Thus we predict
the possibility of observing quantum-like interference for macroscopic objects.
By using the contextual probabilistic calculus developed in this chapter we
can associate complex wave-amplitudes with macroscopic objects. This ap-
proach eliminates the gap between micro and macro worlds.
M. Macroscopic quantum computers
The above arguments give the possibility of creating experimental situa-
tions in which macro systems will exhibit wave behaviour. This implies the
possibility of creating computing devices based on a quantum-like calculus
of probabilities and composed of macro systems.
N. Macroscopic quantum cryptography
The same arguments imply which, in principle, there can be created
quantum-like cryptographic schemes based on macro systems.
O. Hyperbolic quantum computing and cryptography
Hyperbolic interference can be used (in a way similar to trigonometric in-
terference) to realize hyperbolic quantum-like computing and cryptographic
schemes. On the one hand, hyperbolic quantum-like computing and crypto-
graphic schemes might be realized for super-quantum scales. On the other
hand, they might be realized for macro systems.

213
 P. Cognitive quantum-like models
Our probabilistic derivation of interference rules has been performed in
the general framework. In principle, we can use quantum-like probabilistic
calculus to investigate statistical data obtained in cognitive experiments, see
chapter 8.
[Bohmian Mechanics for Financial Processes] Bohmian Mechanics for
Financial Processes
In this Chapter we use methods of classical and quantum mechanics
for mathematical modeling of price dynamics in the financial market. The
Hamiltonian formalism on the price/price-change phase space is used to de-
scribe the classical-like evolution of prices. This classical dynamics of prices
is determined by ‘hard’ conditions (natural resources, industrial production,
services, and so on). These conditions as well as ‘hard’ relations between
traders in the financial market are mathematically described by the classical
financial potential. In the real financial market ‘hard’ conditions are not the
only source of price changes. The information exchange and market psychol-
ogy play important (and sometimes determining) roles in price dynamics.
We propose to describe this ‘soft’ financial factors by using the pilot wave
(Bohmian) model of quantum mechanics. The theory of financial mental (or
psychological) waves is used to take into account market psychology. The real
trajectories of prices are determined (by the financial analogue of Newton’s
second law) by two financial potentials: classical-like (‘hard’ market condi-
tions) and quantum-like (‘soft’ market conditions)The financial quantum-like
model presented in this Chapter was created by Olga Choustova [45], a post-
graduate student of the International Center for Mathematical Modeling in
Physics and Cognitive Sciences, University of Växjö. This model is closely
connected with the Bohmian cognitive model presented in Chapter 6..
Since the 1970s the intensive exchange of information in the world of fi-
nances has become one of the main sources of determining the dynamics of
prices. Electronic trading (which became the most important part of the
environment of the major stock exchanges) induces huge information flows
between traders (including the foreign exchange market). Financial contracts
are performed on a new time scale which differs essentially from the old ‘hard’
time scale that was determined by the development of the economic basis of
the financial market. Prices at which traders are willing to buy (bid quotes)
or sell (ask quotes) a financial asset are no longer determined by the con-
tinuous development of industry, trade, services, the situation in the market
of natural resources, and so on. Information (mental, market–psychological)
factors play a very important (and in some situations crucial) role in price
dynamics. Traders performing financial operations work as a huge collective

214
cognitive system. Roughly speaking, the classical-like dynamics of prices (de-
termined) by ‘hard’ economic factors is permanently perturbed by additional
financial forces, mental (or market–psychological) forces, see the book of J.
Soros [177].
In this chapter methods of Bohmian mechanics are used to simulate the
dynamics of prices in the financial market, see [45]. We start with the de-
velopment of the classical Hamiltonian formalism on the price/price-change
phase space to describe the classical-like evolution of prices. This classical
dynamics of prices is determined by ‘hard’ financial conditions (natural re-
sources, industrial production, services, and so on). These conditions, as
well as ‘hard’ relations between traders in the financial market, are mathe-
matically described by the classical financial potential. As we have already
remarked, in the real financial market ‘hard’ conditions are not the only
source of price changes. The information and market psychology play im-
portant (and sometimes determining) roles in the price dynamics. It was
proposed [45] to describe these ‘soft’ financial factors by using the pilot wave
(Bohmian) model of quantum mechanics. The theory of financial mental (or
psychological) waves is used to take into account market psychology. The
real trajectories of prices are determined (by the financial analogue of New-
ton’s second law) by two financial potentials: classical-like (‘hard’ market
conditions) and quantum-like (‘soft’ market conditions).
Such a quantum-like model of financial processes was strongly motivated
by consideration by J. Soros [177] of the financial market as a complex cog-
nitive system. Such an approach he called the theory of reflexivity.reflexivity
In his theory there is a large difference between a market which is ‘ruled’ by
only ‘hard’ economic factors and a market in which mental factors play the
crucial role (even changing the evolution of the ‘hard’ basis, see [177]). J.
Soros rightly remarked that a ‘non-mental market’ evolves during classical
random fluctuations. However, such fluctuations do not provide an adequate
description of the mental market. He proposed to use an analogy with quan-
tum theory. However, it was noticed that quantum formalism could not be
applied directly to the financial market, [177]. Traders differ essentially from
elementary particles. According to J. Soros, traders in the financial market
behave stochastically due to the free will of individuals. Combinations of
a huge number of free wills of traders produce additional stochastics in the
financial market that could not be reduced to classical random fluctuations
(determined by non-mental factors). Here J. Soros followed the conventional
(Heisenberg, Bohr, Dirac) viewpoint of the origin of quantum stochastics.
However, in the Bohmian approach (which is a nonconventional one) quan-
tum statistics is induced by the action of an additional potential, quantum
potential, that changes classical trajectories of elementary particles. Such

215
an approach gives the possibility of applying quantum formalism to the fi-
nancial market. We also remark that recently it was demonstrated that, in
fact, quantum-like (Hilbert space probabilistic) formalism can be applied to
various statistical phenomena outside the micro-world, see Chapters 9, 10.
It seems that the quantum financial approach gives a new possibility of
describing mathematically the stochastics in the financial market — com-
pare with the classical stochastic approach starting with Bachelier’s doctoral
thesis, see also, e.g., [45] for an extended bibliography on the use of different
physical models in the social sciences and the economy.

61 Price Phase Space

Let us consider a mathematical model in which n traders, a1 , . . . , an interact
with one another and react to external economic (as well as political) infor-
mation in order to determine the best price to buy or sell financial assets. We
consider a price system of coordinates. There is the n-dimensional configura-
tion space Q = Rn of prices, q = (q1 , . . . , qn ), where qj is the price proposed
by the jth trader. Here R is the real line. The dynamics of the prices is
described by the trajectory q(t) = (q1 (t), . . . , qn (t)) in the configuration price
space Q.
The reader may be surprised that we plan to use the whole real line R
(and not just the positive half-line) to describe price dynamics for a trader.
Well, it is clear what is the meaning of the price q = 1 dollar. But: What
is the meaning of the price q = −1 dollar? We shall use the following
interpretation. If trader aj is selling a product (ask quotes), then qj ≥ 0. If
aj is buying a product (bid quotes), then qj ≤ 0.
Another variable under consideration is the price change variable:
qj (t + ∆t) − qj (t)
vj (t) = q̇j (t) = lim .
∆t→0 ∆t
In real models we consider the discrete time scale ∆t, 2∆t, . . . . Here we shall
use the discrete price change variable

zj (t) = qj (t + ∆t) − qj (t).

We denote the space of price changes by the symbol V (≡ Rn ), v = (v1 , . . . , vn ).

As in classical physics, it is useful to introduce the phase space Q × V = R2n ,
namely the price phase space.
A pair (q, v) = (price, price change) is called a state of the financial
market.Later we shall consider quantum-like states of the financial market.
A state (q, v) is a classical state.

216
 We now introduce an analogue m of mass as the number of items (i.e.,
options) which a trader presents to the market‘Number’ is typically a natural
number m = 0, 1, . . . , because even in trade for ‘continuous products’ (such
as oil or gas) we use discrete units, i.e., ton or barrel.. We call m the financial
mass. financial mass Thus each trader has his own financial mass mj . The
total price of his offer to the market is equal to Tj = mj qj .
We also introduce the financial energyfinancial energy of trade as a func-
tion H : Q × V → R. If we use the analogue with classical mechanics Why
not? In principle, there is not much difference between motions in ‘physical
space’ and ‘price space’. then we can consider (at least for mathematical
modeling) the financial energy in the form:
n
1X
H(q, p) = mj vj2 + V (q1 , . . . , qn ). (155)
2 j=1

Here K = 21 nj=1 mj vj2 is the financial kinetic energyfinancial kinetic energy

P
and V (q1 , . . . , qn ) is the financial potential energy, and mj is the financial
mass of jth trader.
The financial kinetic energy represents efforts of traders to change prices:
higher price changes vj induce higher financial kinetic energies. The financial
mass also plays an important role: if one trader a1 sells 1 item and another
trader a2 sells 2 items, and they both change the price in the same way, then
a2 has a financial kinetic energy two times larger than a1 . We remark that
the financial kinetic energy does not depend upon the absolute magnitudes of
prices (only upon price changes). We also remark that high financial kinetic
energy induces rapid changes in the financial situation in market. However,
the financial kinetic energy does not give the attitude of these changes. It
could be rapid economic growth as well as recession.
The financial potential energyfinancial potential energy V describes the
interactions between traders a1 , . . . , an as well as external economic condi-
tions. For example, we can consider the simplest interaction potential:
n
X
V (q1 , . . . , qn ) = (qi − qj )2 .
j=1

The difference |q1 − qj | between the prices of ai and aj is the most im-
portant condition for arbitrage.arbitrage So the description of interactions
between different traders by using the financial potential energy is straight-
forward. What about the role of external conditions? It is a very complicated
problem. In some sense V describes (if we forget about interactions between
the traders a1 , . . . , an ) the reactions of our traders to external conditions.

217
There is a large variety of such conditions, economic as well as political.
For example, suppose that we study the car market. Here a1 , . . . , an are car
traders (selling as well as buying cars). Then the financial potential energy
depends, in particular, on the oil price (which influences car prices)
We can never take into account all economic and other conditions that
have influences on the market. Therefore by using some concrete potential
V (q) we consider the very idealized model of financial processes. However,
such an approach is standard for physical modeling where we also consider
idealized mathematical models of real physical processes.
To describe the dynamics of prices it is natural to use Hamiltonian dynam-
ics on the price phase space. As in classical mechanics for material objects, it
is useful to introduce a new variable p = mv, the price momentumprice mo-
mentum variable. So instead of the price change vector v = (v1 , . . . , vn ) we
shall consider the price momentum vector p = (p1 , . . . , pn ), pj = mj vj . The
space of price momenta is denoted by the symbol P. The space Q × P will
also be called the price phase space. The Hamiltonian equations of motion
on the price phase space have the form:
∂H ∂H
q̇ = , ṗj = − , j = 1, . . . , n. (156)
∂pj ∂qj
We also need the initial conditions, qj (0) = qj0 , pj (0) = pj0 . If the financial
energy has form (155), i.e.,
n
X p2j
H(q, p) = + V (q1 , . . . , qn ),
j=1
2m j

then the Hamiltonian equations have the form

pj ∂V
q̇j = = vj , ṗj = − .
mj ∂qj
The latter equation can be written in the form:
∂V
mj v̇j = − .
∂qj
It is natural to call the quantity
vj (t + ∆t) − vj (t)
v̇j = lim
∆t→0 ∆t
the price accelerationprice acceleration (the change of price change). The
quantity
∂V
fj (q) = −
∂qj

218
is called the (potential) financial force. We obtain the financial variant of
Newton’s second law:
mv̇ = f. (157)
The product of the financial mass and the price acceleration is
equal to the financial force.
In fact, the Hamiltonian evolution is determined by the following funda-
mental property of the financial energy: The financial energy is not changed
in the process of Hamiltonian evolution:

H(q1 (t), . . . , qn (t), p1 (t), . . . , pn (t) = H(q1 (0), . . . qn (0), p1 (0), . . . , pn (0))

We need not restrict our considerations to financial energies of the form (155).
First of all external (e.g., economic) conditions as well as the character of
interactions between traders in the market depend strongly on time. This
must be taken into account by considering time-dependent potentials, V =
V (t, q). Moreover, the assumption that the financial potential depends only
on prices, V = V (t, q), in not as natural for the modern financial market.
Traders have complete information on price changes. This information is
taken into account by traders for acts of arbitrage. Therefore it can be useful
to consider potentials which depend not only on prices, but also on price
changes, V = V (t, q, v), or, in the Hamiltonian framework, V = V (t, q, p).
In such a case the financial force is not potential. Therefore it is also useful
to consider the financial Newton’s second law for general financial forces:
mv̇ = f (t, q, p).
Remark 1.1. (On the form of the financial kinetic energy). We copied
the form of kinetic energy from classical mechanics for material objects. It
may be that such a form of financial kinetic energy is not justified by the real
financial market. It might be better to consider our choice of the financial
kinetic energy as just the basis of mathematical modeling (and looking for
other possibilities).
Example 1.1. (Free trader) Let n = 1 and V ≡ 0. We obtain ṗ = 0 or
p(t) ≡ p0 and q̇ = pm0 = v0 . Thus q(t) = q0 + v0 t. Let v0 > 0. So in the absence
of concurrence and by neglecting external (economic and political conditions)
a trader will increase the price linearly. It is very natural behaviour If the
market has very large capacity. However, in this example infinite capacity
of a market is a consequence of the condition V ≡ 0 : there are no external
constraints..
We now consider the situation in which v0 < 0. Here a trader will linearly
decrease a price. If at t = t0 a trader (on the basis of some information from
market) decided to decrease the price and after this he can not obtain any

219
information on economic and financial situation. So he continues to decrease
the price. Of course, this is a very idealized example. We could never find an
isolated trader. Any trader needs a partner for arbitrage, and consequently
can use information about a partner’s behaviour.

62 Hamiltonian Price Dynamics and the Stock

Market
The model of Hamiltonian price dynamics on the price phase space can be
useful for describing a market which essentially depends on ‘hard’ economic
conditions: natural resources, volumes of production, human resources, .... In
principle, such a model can even be used in a plan economy: by introducing
different potentials V (q1 , . . . , qn ) we can regulate ‘plan–market’, see, e.g., [45].
However, it seems that classical price dynamics can not be applied (at least
directly) to financial markets [177]. Well, ‘hard’ economic conditions play an
important role in forming stock prices. However, it is clear that the stock
market is not based only on these ‘hard’ factors. There are other factors,
soft ones, which play an important and (sometimes even determining) role in
forming prices in the financial market. We can not define these soft factors
precisely. We can speak about a market’s psychology. These psychological
factors became very important during the rapid exchanges of information
performed by modern financial computer systems.
Negligibly small amounts of information (owing to the rapid exchange
of information) can imply large changes of prices in the financial market.
We may speak about financial (psychological) waves that are permanently
present in the financial market. Sometimes these waves produce uncontrol-
lable changes of prices disturbing the whole financial market (financial crises).
Of course, financial waves depend on hard economic factors. However, these
hard factors do not play the crucial role in forming of financial waves. Finan-
cial waves are merely waves of information. We can again use the example of
Bohm–Hiley which has already been considered in Chapter 6. We can com-
pare the behaviour of the financial market with the behaviour of a gigantic
ship which is controlled by a radio signal. A radio signal with negligibly
small physical energy can essentially change (owing to information contained
in the signal) the motion of the gigantic ship. If we do not pay attention
to (do not know about the presence of) the radio signal, then we will be
continuously disappointed by ship’s behaviour. It can change its direction of
motion without any ‘hard’ reason (weather, destination, the technical state
of the ship’s equipment). However, if we know about the existence of radio

220
monitoring, then we can find the information which is sent by radio. This
would give us a powerful tool for predicting ship’s trajectory.

63 Financial Pilot Waves

According to the Bohmian interpretation of quantum mechanics a quantum
system consists of a material body (e.g., an elementary particle) and a pilot
wave. The latter ‘controls’ the first.
If we interpret the pilot wave as a kind of physical field then we should
recognize that this is a rather strange field. It differs crucially from other
physical fields, e.g., the electromagnetic field. The latter carries physical
energy. There are some other pathological features of the pilot wave field.
In particular, the force induced by this pilot wave field does not depend on
the amplitude of wave. Thus small waves and large waves change equally
the trajectory of an elementary particle. Such features of the pilot wave
give the possibility of speculating, see Chapter 6, that this is just a wave
of information (active information). Hence the pilot wave field does not
describe the propagation of energy in physical space-time but the propagation
of information. The pilot wave is more similar to a radio signal which guides
a ship. Of course, this is just an analogy (because a radio signal is related to
an ordinary physical field, namely, electromagnetic field). The more precise
analogy is to compare the pilot wave with information contained in a radio
signal.
We remark that the pilot wave (the Bohmian) interpretation of quantum
mechanics is not the conventional one. As we have already noted, there are
a few critical arguments against the Bohmian quantum formalism:
1. Bohmian theory gives the possibility of providing a mathematical
description of the trajectory q(t) of an elementary particle. However, such a
trajectory does not exist according to the conventional quantum formalism.
2. Bohmian theory is not local, namely, via the pilot wave field one
particle ‘feels’ another at large distances (without any exchange of physical
energy).
Regarding 1 and 2 we say that these disadvantages of the theory will
become advantages in our applications of Bohmian theory to the financial
market.
Our fundamental assumption is that traders in the modern financial mar-
ket are not just classical-like traders. Their actions are controlled not only
by classical-like financial potentials V (t, q1 , . . . , qn , p1 , . . . , pn ) but also (in the
same way as in the pilot wave theory for quantum systems) by an additional

221
information (or psychological, compare [177]) potential induced by a financial
pilot wave.
Therefore we can not use the classical financial dynamics (the Hamil-
tonian formalism) on the financial phase space to describe the real price
trajectories. Information (psychological) perturbation of the Hamiltonian
equations for price and price change must be taken into account. To describe
such a model mathematically, it is convenient to use an object such as a
financial pilot wavefinancial pilot wave which controls the activity of each
trader in the financial market.
The reader may ask: ‘Where is such a financial field defined?’ In principle,
it is possible to develop a model in which this field is distributed over phys-
ical space (the surface of the Earth with singularities at New-York, Tokyo,
London, Paris, Frankfurt, ...). However, we prefer to use information space,
namely, the price space Q, in our modeling. Thus the financial pilot wave
is mathematically described as a function ϕ : Q → C, where C is the set
of complex numbers. It maps the price configuration q = (q1 , . . . , qn ) for n
traders a1 , . . . , an into a complex number ϕ(q), the amplitude of the price
configuration. In some sense ϕ(q) describes the psychological influence of
the price configuration q on traders. The reader may be surprised that the
complex numbers C appeared. However, the use of these numbers is just a
mathematical trick which provides a simple mathematical description of the
dynamics of the financial pilot wave. Before going into mathematical details
we underline two important features of the financial pilot wave model:
1. All traders are coupled at the information level. The general formal-
ism of the pilot wave theory says that if the function ϕ(q1 , . . . , qn ) is not
factorized, i.e.,
ϕ(q1 , . . . , qn ) 6= ϕ1 (q1 ) . . . ϕn (qn ),
then by changing the price qi the trader ai will automatically change the
behaviour of all the other traders aj , j 6= i. At the same time the ‘hard’
economic potential V (q1 , . . . , qn ) can be totally local: economic conditions
for distinct traders can be independent. For example, V (q1 , . . . , qn ) = q12 +
. . . qn2 and a1 , . . . , an are totally independent economically. The Hamiltonian
equations in the absence of the financial pilot wave have the form: q̇j =
pj , ṗj = −2qj , j = 1, 2, . . . , n. Thus the classical-like price trajectory qj (t), of
aj , does not depend on the dynamics of prices for other traders ai , i 6= j.
However, if, e.g.,
2 2
ϕ(q1 , . . . , qn ) = cei(q1 q2 +...+qn−1 qn ) e−(q1 +...+qn ) ,
where c ∈ C is some normalization constant, then the financial behaviour
of traders in the financial market is nonlocal (see further considerations).

222
Everybody would immediately react to changes of other prices, despite a
stable domestic economic situation.
2. The reactions of traders aj do not depend on the amplitude of the
financial pilot wave: financial waves ϕ, 2ϕ, 100000ϕ will produce the same
reactions of aj . Such a behaviour of traders in the financial market is quite
natural (if the financial pilot wave is interpreted as an information wave, the
wave of financial information). The amplitude of an information signal does
not play as large a role in information exchange. The most important is the
context of such a signal. The context is given by the shape of the signal, the
form of the financial pilot wave function.

64 The Dynamics of Prices Guided by the Fi-

nancial Pilot Wave
In fact, we need not develop a new mathematical formalism. We will just
apply the standard pilot wave formalism to traders in the financial market.
The fundamental postulate of the pilot wave theory is that the pilot wave
(field) ϕ(q1 , . . . , qn ) induces a new (quantum) potential U (q1 , . . . , qn ) which
perturbs the classical equations of motion. A modified Newton equation has
the form:
ṗ = f + g,
where f = − ∂V ∂q
and g = − ∂U ∂q
. We call the additional financial force g a
financial mental force.financial mental force This force g(q1 , . . . , qn ) deter-
mines a kind of collective consciousness of the financial market. Of course,
the g depends on economic and other ‘hard’ conditions given by the financial
potential V (q1 , . . . , qn ). However, this is not a direct dependence. In prin-
ciple, a nonzero financial mental force can be induced by the financial pilot
wave ϕ in the case of zero financial potential, V ≡ 0. So V ≡ 0 does not
imply that U ≡ 0. Market psychology is not a totally determined by economic
factors. Financial (psychological) waves of information need not be gener-
ated by some changes in a real economic situation. They are mixtures of
mental and economic waves. Even in the absence of economic waves, mental
financial waves can have a large influence to the financial market.
By using the standard pilot wave formalism we obtain the following rule
for computing the financial mental force. We represent the financial pilot
wave ϕ(q) in the form:
ϕ(q) = R(q)eiS(q)
where R(q) = |ϕ(q)| is the amplitude of ϕ(q), (the absolute value of the
complex number c = ϕ(q)) and S(q) is the phase of ϕ(q) (the argument

223
of the complex number c = ϕ(q)). Then the financial mental potential is
computed as
n
1 X ∂ 2R
U (q1 , ..., qn ) = −
R i=1 ∂qi2
and the financial mental force as
−∂U
gj (q1 , . . . , qn ) = (q1 , . . . , qn ).
∂qj
These formulas imply that strong financial effects are produced by financial
waves having essential variations of amplitudes.
Example 4.1. (Financial waves with small variation have no effect). Let
R ≡ const. Then the financial mental force g ≡ 0. As R ≡ const the jth
trader can not perturb the market’s consciousness by varying his price qj .
The constant information field does not induce psychological financial effects
at all. As we have already remarked the absolute value of this constant does
not play any role. Waves of constant amplitude R = 1, as well as R = 10100 ,
produce no financial effect.
Let R(q) = cq, c > 0. This is a linear function; variation is not so large.
As the result g ≡ 0 here also. No financial mental effects.
Example 4.2. (Successive speculations) Let R(q) = c(q 2 + d), c, d > 0.
Here U (q) = − q22+d (it does not depend on the amplitude c !) and g(q) =
−4q
(q 2 +d)2
. The quadratic function varies essentially more strongly than the linear
function, and, as a result, such a financial pilot wave induces a nontrivial
financial mental force.
We analyse financial drives induced by such a force for traders selling
options. Here q > 0 and g < 0. The financial mental force g stimulates
trader a to decrease the price. For small prices, g(q) ≈ −4q/d2 . If trader a
increases his price q for options then the negative reaction of the financial
mental force becomes stronger and stronger. He is pressed by the financial
market to stop increase of the price q. However, for large prices, g(q) ≈ −4/q 3 .
If the a can approach this range of prices (despite the negative pressure of
the financial market for relatively small q) then the a will feel decreasing
of the negative pressure of the financial market. This model explains well
successive speculative behaviour in the financial market.
5
Let R(q) = c(q 4 + b), c, b > 0. Thus g(q) = (qbq−q4 +b)2 . Here behaviour of

trader a is√more complicated. We consider the case q ≥ 0 (asking quotes).

Let d =4 b. If the price q is changing from q = 0 to q = d then the a
is motivated (by the financial mental force g(q)) to increase the price. The
price q = d is critical for his financial activity. For psychological reasons (of

224
course, indirectly based on the whole information on the financial market)
he understands that it would be dangerous to continue to increase the price.
Since q = d he has the psychological stimuli to decrease the price.
Financial pilot waves ϕ(q) with R(q) that are polynomials of higher order
can induce very complex behaviour. The interval [0, ∞) is split into a collec-
tion of subintervals 0 < d1 < d2 < . . . < dn < ∞ such that at each price level
q = dj the trader changes his attitude to increase or to decrease the price.
In fact, we have considered just a one-dimensional model. In the real
case we have to consider multidimensional models of huge dimension. A
financial pilot wave ϕ(q1 , . . . , qn ) on such a price S Q induces splitting
S space
of Q into a large number of domains Q = O1 . . . ON . Each domain Oj
is characterized by a fixed collection of financial attitudes of traders in the
financial market. For example, in O1 : a1 has the drive to increase the price
(and he is selling options, q1 ≥ 0), a2 has the drive to decrease the price (and
he is buying options q2 ≤ 0), ..., an has the drive to decrease the price (and
he is selling options qn ≥ 0).
The only problem which we have still to solve is the description of the
time-dynamics of the financial pilot wave, ϕ(t, q). We follow the standard
pilot wave theory. Here ϕ(t, q) is found as the solution of Schrödinger’s
equation. The Schrödinger equation for the energy
n
1 X p2j
H(q, p) = + V (q1 , . . . , qn )
2 j=1 mj

has the form:

∂ϕ
ih (t, q1 , . . . , qn ) =
∂t
n
X h2 ∂ 2 ϕ(t, q1 , . . . , qn )
− + V (q1 , . . . , qn )ϕ(t, q1 , . . . , qn ), (158)
j=1
2mj ∂qj2

with the initial condition ϕ(0, q1 , . . . , qn ) = ψ(q1 , . . . , qn ).

Thus if we know ϕ(0, q) then by using Schrödinger’s equation we can
find the pilot wave at any instant of time t, ϕ(t, q). Then we compute the
corresponding mental potential U (t, q) and mental force g(t, q) and solve
Newton’s equation.
We shall use the same equation to find the evolution of the financial pilot
wave. We have only to make one remark, namely, on the role of the con-
stant h in Schrödinger’s equation. In quantum mechanics (which deals with
microscopic objects) h is the Planck constant. This constant is assumed to
play the fundamental role in all quantum considerations. However, originally

225
h appeared as just a scaling numerical parameter for processes of energy ex-
change. Therefore in our financial model we can consider h as a price scaling
parameter, namely, the unit in which we would like to measure price change.
We do not present any special value for h. There are numerous investigations
into price scaling. It may be that there can be recommended some special
value for h related to the modern financial market, a fundamental financial
constant.fundamental financial constant However, it seems that h = h(t)
evolves depending on economic development.
We suppose that the financial pilot wave evolves via the financial Schrödinger
equation (an analogue of Schrödinger’s equation) on the price space. In the
general case this equation has the form:
∂ϕ
ih (t, q) = Hϕ(t,
b q), ϕ(0, q) = ψ(q),
∂t
where H b is self-adjoint operator corresponding to the financial energy given
by a function H(q, p) on the financial phase space. Here we have to proceed
in the same way as in ordinary quantum theory for elementary particles.
As the mathematical basis of the model we use the space L2 (Q) of square
integrable functions ϕ : Q → C, where Q is the configuration price space,
Q = Rn , or some domain O ⊂ Rn :
Z
2
||ϕ|| = |ϕ(x)|2 dx < ∞.
Q

Here dx is the Lebesque measure, a uniform probability distribution, on

the configuration price space.
Of course, the uniform distribution dx is not the unique choice of the
normalization measure on the configuration price space. By choosing dx we
assume that in the absence of the pilot wave influence, e.g., for ϕ(x) ≡ const,
all prices ‘have equal rights’. In general, this is not true. If there is no
financial (psychological) waves the financial market still strongly depends
on ‘hard’ economic conditions. In general, the choice of the normalization
measure M must be justified by a real relation between prices. So in general
the financial pilot wave ϕ belongs to the space L2 (Q, dM ) of square integrable
functions with respect to some measure M on the configuration price space:
Z
2
||ϕ|| = |ϕ(x)|2 dM (x) < ∞.
Q

In particular, M can be a Gaussian measure:

1 −(B −1 (x−α),x−α)
dM (x) = e 2 dx,
(2πdetB)n/2

226
where B = (bij )ni,j=1 is the covariance matrix and α = (α1 , . . . , αn ) is the av-
erage vector. The parameters (bij ) and α are determined by ‘hard’ economic
conditions. In particular, bij gives the covariation between the prices of the
traders ai and aj We recall that the matrix B is symmetric. Thus bij = bji .
The αj is the average of prices aj , αj = q̄j , proposed by the trader aj . The
measure M describes classical random fluctuations in the financial market
that are not related to ‘quantum’ effects. The latter effects are described in
our model by the financial pilot wave. If the influence of this wave is very
small we can use classical probabilistic models; in particular, based on the
Gaussian distribution. The Gaussian model for price fluctuations was the
first financial probabilistic model — Bachelier’s model. In fact, Bachelier
described price dynamics by Brownian motion. Therefore it would be even
more natural to consider the Gaussian distribution of price changes. So it
is useful to the study momentum representation for the pilot wave theory
which was recently developed by B. Hiley, see [121]. Instead of financial
waves on the configuration financial space Q we can consider financial waves
ϕ : P → C on the momentum space. We recall that the momentum p = mv,
where v, the velocity, describes price changes. Therefore by following Bache-
lier we have to consider the Gaussian representation, ϕ ∈ L2 (P, dm), where
P is the price change space.
However, further investigations demonstrated that it seems that the Gaus-
sian model for price changes is not the best one for describing price fluctua-
tions. One of alternative models is based on the Levy process. Therefore it
can be useful to investigate the ‘quantum’ financial model that is based on
the Cauchy (Lorentzian) measure:
γ dp
dM (p) =
π γ 2 + p2
on the momentum financial space. It would be interesting to provide compar-
ative analysis of financial pilot wave models for dM (p) Gaussian and dM (p)
Cauchy.
It seems that the Cauchy measure has some advantages. By using this
measure we exclude from consideration financial waves which increase at
infinity as a linear function. So ϕ(p) ≈ pk , k ≥ 1, p → ∞, are excluded from
this model. It looks very natural since prices (at least in the real financial
market) can not change arbitrarily quickly.
We now turn back to the general scheme, concentrating on the configura-
tion representation, ϕ : Q → C; ϕ ∈ L2 (Q) ≡ L2 (Q, dx). This is the general
quantum-like statistical formalism on the price space.
As in ordinary quantum mechanics, we consider a representation of fi-
nancial quantities, observables, by symmetric operators in L2 (Q). By using

227
Schrödinger’s representation we define price and price change operators by
setting:
q̂j ϕ(q) = qj ϕ(q),
the operator of multiplication by the qj -price;

h ∂
p̂j = ,
i ∂qj

the operator of differentiation with respect to the qj -price, normalized by the

scaling constant h (and −i = 1i which provides the symmetry of p̂j ). Opera-
tors of price and price change satisfy the canonical commutation relations:

[q̂, p̂] = q̂ p̂ − p̂q̂ = ih.

By using this operator representation of price and price changes we can rep-
resent every function H(q, p) on the financial phase-space as an operator
H(q̂, p̂) in L2 (Q). In particular, the financial energy operator is represented
by the operator:
n
X h2 ∂ 2
Ĥ = − + V (q1 , . . . , qn ).
j=1
2mj ∂qj2

Here V (q) is the multiplication operator by the function V (q).

In this general quantum-like formalism for the financial market we do not
consider individual evolution of prices. The theory is purely statistical. We
can only determine the average of a financial observable A for some fixed
state φ of the financial market:
Z
hAiφ = A(φ)(x)φ̄(x)dx.
Q

The use of the Bohmian model gives the additional possibility of determining
individual trajectories.

228
 4

1. Akiskal, H.S., McKinney, W.T. (1973). ‘Depressive disorders: toward

a unified hypothesis’, Science 182, 20–29.

2. Albeverio, S., Khrennikov, A.Yu. and Cianci, R. (1997), ‘On the spec-
trum of the p-adic position operator’, J. Phys. A, Math. Gen., 30,
881–889.

3. Albeverio, S., Khrennikov, A.Yu., De Smedt, S. and Tirozzi, B. (1998).

‘p-adic dynamical systems’, Theor. Math. Phys., 114, 349–365.

4. Albeverio, S., Khrennikov, A.Yu. and Kloeden, P. (1999). ‘Memory

retrieval as a p-adic dynamical system’, Biosystems, 49, 105–115.

5. Albeverio, S, Khrennikov, A Yu. and Tirozzi, B. (1999). ‘p−adic

Neural Networks’, Math. Models Meth. Appl. Sci., 9, 1417–1437.

6. Albeverio, S., Gundlach, M., Khrennikov, A.Yu., Lindahl, K.-O. (2001).

‘On Markovian behaviour of p-adic random dynamical systems’. Russ.
J. Math. Phys., 8(2), 135–152.

7. Alcock, J.E. (1981). Parapsychology: Science or Magic? (Pergamon

Press: Elmsford, NY).

8. Alcock, J.E. (1987). ‘Parapsychology: science of the anomalous or

search for the soul’, Behav. Brain Sci., 10, 553–565.

9. Allen, W.F. (1940). ‘Effect of ablating the frontal lobes, hippocampi,

and occipito-parieto-temporal (excepting pyriform areas) lobes on pos-
itive and negative olfactory conditional reflexes’, Amer. J. Physiol.,
128, 754–771.

10. Allport, D.A. (1977). ‘On knowing the meaning of words we are unable
to report: The effects of visual masking’, (Ed. Dornic S.,) Attention
and Performance., 6, 505–533 (Erlbaum, Hillsdale N. J.).

11. Amit, D.J. (1989). Modeling of Brain Functions (Cambridge University

Press: Cambridge).

12. Amit, D.J. (1990). ‘Attractor neural networks and biological reality:
associative memory and learning’, Future Generation Comput. Sys.,
6, 111–119.

229
13. Aref’eva, I.Ya., Dragovic, B. and Volovich, I.V. (1988). ‘On the p-
adic summability of the anharmonic oscillator’, Phys. Lett. B, 200,
512–514.

14. Aref’eva, I.Ya., Dragovich, B., Frampton, P.H., Volovich, I.V. (1991).
‘The wave function of the Universe and p-adic gravity’, Int. J. Mod.
Phys. A, 6, 24, 4341–4358.

15. Arnold, L. (1998). Random Dynamical Systems (Springer–Verlag: Berlin,

New York, Heidelberg).

16. Arrowsmith, D.K. and Vivaldi, F. (1993). ‘Some p−adic representa-

tions of the Smale horseshoe’, Phys. Lett., A 176, 292–294.

17. Arrowsmith, D.K. and Vivaldi, F. (1994). ‘Geometry of p−adic Siegel

discs’, Physica D, 74, 222–236.

18. Ashby, W.R. (1940). ‘Adaptiveness and equilibrium’, J. Mental Sci.,

86, 478–483.

19. Baars, B.J. (1997). In the Theater of Consciousness. The Workspace

of Mind (Oxford University Press: Oxford).

20. Baddeley, A.D., Hitch, G. (1977). ‘Working memory’, (Ed. G.H.

Bower) The psychology of learning and motivation, 80, 47–90, (Aca-
demic Press: London).

21. Ballentine, L.E. (1989). Quantum mechanics (Prentice Hall: Engle-

wood Cliffs, NJ).

22. Ballentine, L.E. (1970). ‘The statistical interpretation of quantum me-

chanics’, Rev. Mod. Phys., 42, 358–381.

23. Balota, D.A. (1983). ‘Automatic semantic activation and episodic

memory encoding’, J. Verb. Learning Verbal Behavior., 22, 88–104.

24. Barrett, J.A. (1999). The quantum mechanics of minds and worlds
(Oxford University Press: Oxford).

25. Beardon, A.F. (1991). Iterations of rational functions (Springer–Verlag:

Berlin, New York, Heideberg).

26. Bechterew, W. (1911). Die Funktionen der Nervencentra (Fischer:

Jena).

230
27. Bell, J. (1987). Speakable and unspeakable in quantum mechanics (Cam-
bridge University Press: Cambridge).
28. Bergson, H. (1896). Matire et Memoire (Alcan: Paris).
29. Bergson, H. (1919). L’Energie spirituelle (Alcan: Paris).
30. Bianchi, L. (1895). ‘The functions of the frontal lobes’, Brain, 18,
497–530.
31. Birkhoff, G. (1967). Lattice theory (American Mathematical Society:
Providence, RI).
32. Bogolyubov, N.N. and Shirkov, D.V. (1984). Introduction to the theory
of quantized fields (Nauka: Moscow).
33. Bogolyubov, N.N., Logunov, A.A., and Todorov, I.T. (1969). Fun-
damentals of axiomatic approach to the quantum field theory (Nauka:
Moscow).
34. Bohm, D. (1951). Quantum theory (Prentice-Hall: Englewood Cliffs,
NJ).
35. Bohm, D. (1987). Unfolding meaning (Routledge & Kegan Paul: Lon-
don).
36. Bohm, D. (1980). Wholeness and the implicate order (Routledge &
Kegan Paul: London).
37. Bohm, D. and Hiley, B. (1993). The undivided universe: an ontolog-
ical interpretation of quantum mechanics (Routledge & Kegan Paul:
London).
38. Boole, G. (1847). The mathematical analysis of logic, being an essay
towards a calculus of deductive reasoning (Cambridge University Press:
Cambridge). Reprinted, Philosophical Library, New York, 1948.
39. Boole, G. (1854). An investigation of the laws of thought, on which
are founded the mathematical theories of logic and probabilities (Walter
and Maberly: London).
40. Borevich, Z.I. and Shafarevich, I.R. (1966). The number theory (Aca-
demic Press: New York).
41. Broadbent, D.E. (1958). Perception and communication (Pergamon:
New York).

231
42. Broadbent, D. E. (1982). ‘Task combination and selective intake of
information’, Acta Psych., 50, 253–290.

43. Cartan, H. (1967). Calcul différentiel. Formes différentielles (Her-

mann: Paris).

44. Chalmers, D.J. (1996). The conscious mind: in search of a fundamental

theory (Oxford University Press: New York).

45. Choustova, O.A. (2001). ‘Pilot wave quantum model for the stock mar-
ket’, Quantum theory: reconsideration of foundations. Proceedings of
the International Conference in Växjö, June 2001. Ed. A. Khrennikov.
Series Mathematical Modelling in Physics, Engineering and Cognitive
sciences, 1, 41-58 (Växjö University Press: Växjö).

46. Clark, A. (1980). Psychological models and neural mechanisms. An

examination of reductionism in psychology (Clarendon Press: Oxford).

47. Cohen, J.D. , Perlstein, W.M., Braver, T.S., Nystrom, L.E., Noll, D.C.,
Jonides, J. and Smith, E.E. (1997). ‘Temporal dynamics of brain acti-
vation during working memory task’, Nature, 386, April 10, 604–608.

48. Collewijn, H., Winterson, B.J., Dubois, M.F.W. (1978). ‘Optokinetic

eye movements in albino rabbits: inversion in anterior visual field’,
Science, 199, 1351–1353.

49. Compton, A.H. (1953). The freedom of man (Greenwood Press: NY).

50. Courtney, S.M., Ungerleider, L.G., Keil, K. and Haxby J.V. (1997).
‘Transient and susteined activity in a disturbed neural system for hu-
man working memory’, Nature, 386, April 10, 608–611.

51. Craik, F.I.M. and Lockhart, R.S. (1972). ‘Levels of processing: A

framework for memory research’, J. Verb. learning and Verb. Behav.,
11, 671–684.

52. Damasio, A.R. (1994). Descartes’ error: emotion, reason, and the
human brain (Anton Books: NY).

53. Damasio, H. and Damasio, A.R. (1989). Lesion analysis in neuropsy-

chology (Oxford University Press: New York).

54. Dawkins, R. (1976). The selfish gene (Oxford University Press: New
York).

232
55. Dawson, G.D. (1951). ‘A summation technique for detecting small
signals in a large irregular background’, J. of Physiology, 115, 2–3.

56. De Broglie, L. (1964). The current interpretation of wave mechanics.

A critical study (Elsevier: Amsterdam).

57. de Groot, J. (1956). ‘Non-Archimedean metrics in topology’, Proc.

Amer. Math. Soc., 7(6), 948–956.

58. Dennet, D.C. (1991). Consciousness explained (Little, Brown & Com-
pany: Boston, Toronto, London).

59. Descartes, R. (1680). Six metaphysical meditations; wherein it is proved

that there is a God and that man’s mind is really distinct from his body
(W. Molyneux: London).

60. d’Espagnat, B. (1999). Conceptual foundations of quantum mechanics

(Perseus Books: Reading, Mass).

61. De Smedt, S. and Khrennikov, A. (1997). ‘Dynamical systems and

theory of numbers’, Comment. Math. Univ. St. Pauli, 46, 2, 117–132.

62. Devaney, R.L (1986). An introduction to chaotic dynamical systems

(Benjamin Cummings: New York).

63. Dirac, P.A.M. (1942). ‘The physical interpretation of quantum me-

chanics’, Proc. Roy. Soc. London A, 180, 1–39.

64. Dirac, P.A.M. (1995). The principles of quantum mechanics (Claren-

don Press: Oxford).

65. Dixon, N. (1981). Preconscious processing (Wiley: New York).

66. Donnellan, T. (1969). Lattice theory (Pergamon Press: Oxford).

67. Dragovic, B. (1995). ‘Adelic wave function of the universe’, Proc. 3rd
A. Friedmann International. Seminar on Grav. Cosmology (St. Pe-
tersburg University: St. Petersburg, Russia).

68. Dubischar, D., Gundlach, V.M., Steinkamp, O. and Khrennikov, A.Yu.

(1999). ‘A p-adic model for the process of thinking disturbed by phys-
iological and information noise’, J. Theor. Biol., 197, 451–467.

69. Dunne, B.J., Bisaha, J.P. (1979). ‘Precognitive remote viewing in the
Chicago area’, J. Parapsychol., 43, 17–30.

233
70. Eccles, J.C. (1974). The understanding of the brain (McGraw-Hill:
New York).

71. Edelman, G. M. (1989). The remembered present: a biological theory

of consciousness (Basic Book Publishers: New York).

72. Engelking, R. (1977). General topology (PWN: Warsaw).

73. Escassut, A. (1995). Analytic elements in p-adic analysis (World Sci-

ence Publishers: Singapure).

74. Feynman, R.P. (1987). ‘Negative probability.: Quantum Implications,

Essays in Honour of David Bohm, B.J. Hiley and F.D. Peat, ed., 235–
246, (Routledge & Kegan Paul: London).

75. Florenskii, P. A. (1990). Pillar and statement of the truth. From the
Russian Philosophical Thought Series, 1 (Pravda: Moscow).

76. Fodor, J.A. (1975). Language of thought (Thomas Y Crowell: New

York).

77. Fodor, J.A. and Pylyshyn, Z. W. (1988). ‘Connectionism and cognitive

architecture: a critical analysis’, Cognition, 280, 3–17.

78. Freud, S. (1933). New introductory lectures on psychoanalysis (Norton:

New York).

79. Freund, P.G.O. and Olson, M. (1987). ‘Non-Archimedean strings’,

Phys. Lett. B, 199, 186–190.

80. Freund, P.G.O. and Witten, E. (1987). ‘Adelic string amplitudes’,

Phys. Lett. B, 199, 191–195 .

81. Fuster, J.M.D. (1997). ‘The prefrontal cortex: anatomy, physiology,

and neuropsychology of the frontal lobe’ (Lippincott-Raven: Philadel-
phia).

82. Galaburda, A.M., LeMay, M., Kemper, T.L. and Geschwind (1978).
‘Right–left asymmetries in the brain’, Science, 199, 852–856.

83. Gratzer, G. (1978). General lattice theory (Birhäuser Verlag: Basel).

84. Gray, J. (1973). The psychology of fear and stress (Weindenfeld and
Nicolson: London).

234
85. Green, M.B., Schwarz, J.H. and Witten, E. (1987). Superstring theory
(Cambridge University Press: Cambridge).

86. Gundlach, M., Khrennikov, A. , Lindahl, K.-O. (2001). ‘Topological

transitivity for p-adic dynamical systems’, p-adic functional analysis.
Ed. A. K. Katsaras, W.H. Schikhof, L. Van Hamme, Lecture notes in
pure and apllied mathematics, 222, 127–132.

87. Gödel, K. (1931). ‘Über formal unentscheidbare Sätse per Principia

Mathematica und verwandter Systeme I’, Monatshefte für Mathematik
und Physik, 38, 173–198.

88. Harkavy, A.A. (1989). ‘Quantum mechanical information is ubiqui-

tous’, (Ed. Kafatos M.) Bell’s theorem, quantum theory and concep-
tions of the universe (Kluwer: Dordrecht).

89. Heisenberg, W. (1930). The physical principles of the quantum theory

(Dover Publications: London).

90. Heisenberg, W. (1958). Physics and philosophy (Harper & Row, Harper
Torchbooks: New York).

91. Hensel, K. (1894). ‘Untersuchung der Fundamentalgleichung einer Gat-

tung für eine reelle Primzahl als Modul und Bestimmung der Theiler
ihrer Discriminante’, J. Reine Angew. Math., 113, 61–83.

92. Hiley, B. and Pylkkänen, P. (1997). ‘Active information and cognitive

science – A reply to Kieseppä. In: Brain, mind and physics. Eds:
Pylkkänen, P., Pylkkö, P., Hautamäki, A. (IOS Press, Amsterdam).

93. Hopfield, J.J. (1982). ‘Neural networks and physical systems with
emergent collective computational abilities’, Proc. Natl. Acad. Sci.
USA, 79, 1554–2558.

94. Hoppensteadt, F.C. (1997). An introduction to the mathematics of

neurons: modeling in the frequency domain (Cambridge Univ. Press:
New York).

95. Hsia, L. (1996). A weak Neuron model with applications to p−adic

dynamical systems. Compositio Math., 100, 227–304.

96. Ingvar, D. H. (1985). ‘ "Memory of the future": an essay on the tem-

poral organization of conscious awareness’, Human Neurobiol., 4, 127–
136.

235
 97. Inhelder, B., Piaget, J. (1958). The growth of logical thinking from
childhood to adolescence (Basic Books: New York).

98. Jacobsen, C.F. (1931). ‘A study of cerebral function in learning: The

frontal lobes’, J. Comp. Neurol., 52, 271–340.

99. Jahn, R.G., Dunne, B. J. (1986). On the quantum mechanics of con-

sciosness, with applications to anomalous phenomena’, Found. Phys.,
16, 721–772.

100. Jibu, M. and Yasue, K. (1984). Quantum brain dynamics and con-
sciousness (J. Benjamins Publishing Company: Amsterdam/Philadelphia).

101. Kant, I (1985). Critique of pure reason (Macmillan Press).

102. Kaplanski, I. (1947). ‘Lattices of continuous functions’, Bull. American

Mathematical Society, 53, 617–622.

103. Kennard, M.A., Spencer, S. and Fountain, G. (1941). ‘Hyperactivity

in monkeys following lesions of the frontal lobes’, J. Neurophysiol., 4,
512–524.

104. Khrennikov, A.Yu. (2001). ‘Quantum-like formalism for cognitive mea-

surements’, (Preprint) quant-ph/0111006.

105. Khrennikov, A.Yu. (1991). ‘p-adic quantum mechanics with p-adic

valued functions’, J. Math. Phys., 32, 932–937.

106. Khrennikov, A.Yu. (1992). ‘Axiomatics of the p-adic theory of proba-

bility’, Dokl. Akad. Nauk, 326, 5, 1075–1079.

107. Khrennikov, A.Yu. (1994). p-adic valued distributions and their ap-
plications to the mathematical physics (Kluwer Academic Publisher:
Dordreht).

108. Khrennikov, A.Yu (2001). (ed.), Foundations of probability and physics.

Proceedings of the International Conference in Växjö, November 2000.
Series PQ-QP: Quantum Probability and White Noise Analysis, 13,
(WSP: Singapore).

109. Khrennikov, A.Yu. (1996). ‘Ultrametric Hilbert space representation

of quantum mechanics with a finite exactness’, Found. Phys., 26, 8,
1033–1054.

236
110. Khrennikov, A.Yu. (1997). ‘The uncertainty relation for coordinate
and momentum operators in the p-adic Hilbert space’, Dokl. Akad.
Nauk, 353, 4, 449–452 (English translation: Dokl. Math., 55, 2, 283–
285).

111. Khrennikov, A.Yu. (1997). Non-Archimedean analysis: quantum para-

doxes, dynamical systems and biological models (Kluwer Academic Pub-
lishers: Dordreht)

112. Khrennikov, A.Yu. (1998). ‘Non-Kolmogorov probabilistic models with

p-adic probabilities and foundations of quantum mechanics’, Stochastic
Analysis and Related Topics 4, 275–304. Series Progress in Probability,
42 (Birkhäuser: Boston, Basel, Berlin).

113. Khrennikov, A.Yu. (1998). ‘Human subconscious as the p-adic dynam-

ical system’, J. Theor. Biol., 193, 179–196.

114. Khrennikov, A.Yu. (1997). Superanalysis, (Nauka: Moscow). English

translation: (1999, Kluwer Academic Publishers: Dordreht).

115. Khrennikov, A.Yu. (1999). ‘Description of the operation of the hu-

man subconscious by means of p-adic dynamical systems’, Dokl. Akad.
Nauk, 365, 458–460.

116. Khrennikov, A.Yu. (1999). ‘Classical and quantum mechanics on in-

formation spaces with applications to cognitive, psychological, social
and anomalous phenomena’, Found. Phys., 29, 1065–1098.

117. Khrennikov, A.Yu. (1999). Interpretations of probability (WSP Inter-

national Science Publishers: Utrecht).

118. Khrennikov, A.Yu. (1999). ‘p-adic information spaces, infinitely small

probabilities and anomalous phenomena’, J. of Scientific Exploration,
4, 13, 665–680.

119. Khrennikov, A.Yu. (2000). ‘p-adic discrete dynamical systems and

collective behaviour of information states in cognitive models’, Discrete
Dynamics in Nature and Society, 5, 59–69.

120. Khrennikov, A.Yu. (2000). ‘Classical and quantum mechanics on p-

adic trees of ideas’, BioSystems, 56, 95–120.

121. Khrennikov, A.Yu (2002). (ed.), Quantum theory: reconsideration of

foundations. Proceedings of the International Conference in Växjö,

237
 June 2001. Series Mathematical Modelling in Physics, Engineering
and Cognitive sciences, 1, (Växjö University Press: Växjö).

122. Khrennikov, A.Yu. and Nilsson M. (2001). ‘On the number of cycles
for p-adic dynamical systems’, J. Numb. Th., 90, 255–264.

123. Khrennikov, A.Yu. (2001). ‘Origin of quantum probabilities’, Pro-

ceedings of the Conference: Foundations of Probability and Physics,
180–200 (World Scientific Publishers, Singapore).

124. Khrennikov, A.Yu. (2001). ‘Linear representations of probabilistic

transformations induced by context transitions’, J. Phys. A: Math.
Gen., 34, 9965–9981.

125. Kogan, I.M. (1968). ‘Information theory analysis of telepathic commu-

nication experimentsť, Radio Eng., 23, 122.

126. Kolmogorov, A.N. (1956). Foundations of the probability theory (Chelsea

Publishing Company: New York).

127. Krasner, M. (1944). ‘Nombres semi-réels et espaces ultramétriques’, C.

R. Acad. Sci. Paris, 219, 433–435.

128. Lacan, J. (1986). Schriften (Quadriga Verlag: Weinheim, Berlin).

129. Lemin, A.J. (2001). ‘The category of ultrametric spaces is isomorphic

to the category of complete, atomic, tree-like, and real graduated lat-
tices LAT’, Algebra universalis, to be published.

130. Lemin, A.J. and Lemin, A.V. (2000). ‘On a universal ultrametric
space’, Topology and its Appl., 103, 339–345.

131. Lindahl, K-O. (1999). ‘On Markovian properties of the dynamics on

attractors of random dynamical systems over p-adic numbers’, Rep.
Växjö University, 8.

132. Lockwood, M. (1989). Mind, brain and quantum (Blackwell, Oxford).

133. Lorenz, K. (1966). On aggression (Harcourt, Brace and World: New

York).

134. Lossky, N.O. (1915). ‘The world as the organic whole’, Questions of
Phil. and Psych., 126–129 (in Russian); English translation: (1928),
(Oxford University Press: Oxford).

238
135. Lossky, N.O. (1908). Die Grundlegung des Intuitivismus. Eine propädeutis-
che Erkenntnistheorie (Halle).

136. Lubin, J. (1994). ‘Non-Archimedean dynamical systems’, Compositio

Math., 94, 321–346.

137. Luria, A.R. (1973). The working brain (Penguin books: Harmondsworth,
Middx).

138. Marcel, A.J. (1983). ‘Conscious and unconscious perception: An ap-

proach to the relation between phenomenal experience and perceptual
processes’, Cognitive Psych., 15, 238–300.

139. Marks, D.F. (1986). ‘Investigating the paranormal’, Nature, 320, 119–
124.

140. May, E.C., Humphrey, B.S. and Hubbard, G.S. (1980). ‘Electronic sys-
tem perturbation techniques’, SRI International Final Rep., September
30.

141. Mermin, N.D. (1985).‘Is the moon there when nobody looks? Reality
and quantum theory’, Phys. Today, 38–41, April.

142. Milner, P.M. (1970). Physiological psychology (Holt, Rinehart and Win-
ston).

143. Mises, R. von (1957). Probability, statistics and truth (Macmillan: Lon-
don).

144. Neisser, U. (1967). Cognitive psychology (Appleton-Century-Crofts:

New York).

145. Nilsson, M. (1999). ‘Cycles of monomial dynamical systems over the

field of p−adic numbers.’, Rep. Växjö University, 20.

146. Näätänen, R. (1992). Attention and brain function (Lawrence Erlbaum

Ass. Publ.: Hillsdale, NJ).

147. Oaksford, M. and Chater, N. (1992). ‘Against logicist cognitive sci-

ence’, Mind and Language, 6, 1–38.

148. Orlov, Y.F. (1982). ‘The wave logic of consciousness: A hypothesis’,

Int. J. Theor. Phys., 21, 1, 37–53.

149. Pavlov, I.P. (1949). Complete works (Academy of Science Press: Moscow).

239
150. Penrose, R. (1994). Shadows of the mind (Oxford University Press:
Oxford).

151. Peres, A. (1994). Quantum theory: concepts and methods (Kluwer Aca-
demic Publishers: Dordrecht).

152. Piaget, J. (1952). The Origins of intelligence in children (International

Universities Press, New York).

153. Posner, M.I. and Snyder, C.R.R. (1975). ‘Attention and cognitive con-
trol’, (Ed. Solso R.L. ) Information processing and cognition, 55–85,
The Loyola Symposium (Hillsdale: Erlbaum, N.J).

154. Puthoff, H.E. and Targ, R. (1976). ‘A perceptual channel for infor-
mation transfer over kilometer distances: Historical perspectives and
recent research’, Proc. IEEE, 64, 329–354.

155. Putnam, H. (1994). ‘The best of all possible brains?: Review of shad-
ows of the mind’, The New York Times Book Rev., November 20, p.
1.

156. Pylkkänen, P. (Ed.) (1989). The search for meaning (Wellingborough:

Thorsons).

157. Radin, D.L. and Nelson, R.D. (1989). ‘Evidence for consciousness-
related anomalies in random physical systems’, Found. Phys., 19,
1499–1513.

158. Roland, P.E. (1981). ‘Somatotopical tuning of postcentral gyrus during

focal attention in man. A regional cerebral blood flow study’, J. of
Neurophysiology, 48, 1059–1077.

159. Roland, P. E. (1985). ‘Cortical organization of voluntary behavior in

man’, Human Neurobiol., 4, 155–157.

160. Rolls, E.T. (1975). The brain and reward (Pergamon: Oxford).

161. Ryle, G. (1949). The concept of mind (Barnes & Noble: New York).

162. Sakurai, Y., Sugimoto, S. (1985). ‘Effects of lesions of prefrontal cortex

and dorsomedial thalamus on delayed go/no–go alternative in rats’,
Behav. Brain Res., 17, 213–219.

163. Schikhov, W. (1984). Ultrametric calculus (Cambridge University Press:

Cambridge).

240
164. Schildkraut, J.J. (1973). ‘Neuropharmacology of the affective disor-
ders’, Annual Rev. of Pharmacology, 13, 427–454.

165. Schmidt, H. (1969). ‘Precognition of a quantum process’, J. Parapsy-

chol., 33. 99–108.

166. Schmidt, H. (1970). ‘A PK test with electronic equipment’, J. Para-

psychol., 34, 175–181.

167. Schmidt, H. (1971). ‘Mental influence on random events’, New Sci.

Sci. J., 50, 757–758.

168. Schmidt, H. (1981). ‘PK tests with pre-recorded and pre-inspected seed
numbers’, J. Parapsychol., 45, 87–98.

169. Shannon, C.E. (1948). ‘Mathematical theory of communication’, Bell

System Techn. J., 27, 349–423; 623–656.

170. Schneider, W. and Shiffrin, R.M. (1977). ‘Controlled and automatic

human information processing. I. Detection, search, and attention’,
Psychological Rev., 84, 1–66.

171. Schreiber, H., Lang, M., Lang, W., Kornhuber, A., Heise, B., Keidel,
M., Deecke, L. and Kornhuber, H.H. (1983). ‘Frontal hemispheric dif-
ferences in the Breitschaftspotential associated with writing and draw-
ing’, Human Neurobiology, 2, 197–202.

172. Schrödinger, E. (1992). Philosophy and the birth of quantum mechanics

(Eds. Bitbol, M., Darrigol, O., Editions Frontieres: Gif-sur-Yvette).

173. Seligman, M.E.P. (1975). Helplessness (W.H. Freeman and Co.: San
Francisco).

174. Shiffrin, R.M. and Schneider, W. (1977). ‘Controlled and automatic

information processing: II. Perceptual learning, automatic attending,
and a general theory’, Psychological Rev., 84, 127–189.

175. Shimony, A. (1997). ‘On Mentality, Quantum Mechanics and the Actu-
alization of Potentialities. (Eds. R. Penrose, M. Longair) The large, the
small and the human mind (Cambridge University Press, New York).

176. Skinner, B.F. (1953). Science and human behaviour (Macmillan Co.,
New York).

241
177. Soros, J. (1987). The alchemy of finance. Reading of mind of the
market (J. Wiley and Sons, Inc., New York).

178. Stapp, H.P. (1993). Mind, matter and quantum mechanics (Springer-
Verlag: Berlin-New York-Heidelberg).

179. Sternberg, S. (1966). High speed scanning in human memory, Science,

153, 652–654.

180. Sternberg, S. (1969). ‘The discovery of processing stages: Extension of

Donders’ method’, Acta Psychologica, 30, 276–315.

181. Sternberg, S. (1975). ‘Memory scanning: New findings and current

controversies’, Quaterly J. Exp. Psych., 27, 1–32.

182. Thiran E. and Verstegen, D., Weyers, J. (1989). ‘p−adic dynamics’, J.

Stat. Phys., 54, 893–913.

183. Treisman, A.M. (1960). ‘Contextual cues in selective listening’, Quar-

terly J. Exp. Psych., 12, 242–248.

184. Turing, A.M. (1950). ‘Computing machinery and intelligence’, Mind,

59, 433–460.

185. Umezawa, H. (1993). Advanced field theory: micro, macro, and thermal
physics (American Institute of Physics: New York).

186. Vitiello, G. (2001) My double unveiled - the dissipative quantum model

of brain (J. Benjamins Publ. Company: Amsterdam/Philadelphia).

187. Vladimirov, V.S., Volovich, I.V. and Zelenov, E.I. (1993). p-adic analy-
sis and mathematical physics (World Scientific Publishers: Singapore).

188. Vladimirov, V.S. (1964). Methods of the theory of functions of many

complex variables (Nauka: Moscow).

189. Vladimirov, V.S. and Volovich, I.V. (1989). p-adic quantum mechanics,
Commun. Math. Phys., 123, 659–676.

190. Volovich, I.V. (1987). ‘p-adic string’, Class. Quant. Grav., 4, 83–87.

191. von Neuman, J., (1955). Mathematical foundations of quantum me-

chanics (Princeton University Press: Princeton).

192. Walters, P. (1982). An introduction to ergodic theory (Springer Verlag:

Berlin, New York, Heidelberg).

242
193. Warr, W.B. and Guinan, Jr.J.J. (1979). ‘Efferent innervation of the
organ of Corti: Two separate systems’, Brain Research, 173, 152–155.

194. Wigner, E.P. (1963). ‘The problem of measurement’, Am. J. Phys.,

31, 6.

195. Whitehead A.N. (1929). Process and reality: an essay in cosmology

(Macmillan Publishing Company: New York).

196. Whitehead, A.N. (1933). Adventures of ideas (Cambridge University

Press: London).

197. Whitehead, A.N. (1939). Science in the modern world (Penguin: Lon-
don).

198. Wickens, C.D. (1984). ‘Processing resources in attention’, (Eds. Para-

suraman R. and Davies R.). Varieties of attention, 63–102 (Academic
Press: New York).

199. Wittgenstein, L. (1953). Philosophical investigations (Macmillan: New

York).

200. Wolman, B.B, Ed. (1977). Handbook of parapsychology (Van Nostrand

Reinhold: New York).

201. Zeilinger, A. (1996). ‘On the interpretation and philosophical founda-

tions of quantum mechanics’, Vastakohtien todellisuus. Festschrift for
K.V. Laurikainen. (Ed. Ketvel U. et al.) (Helsinki University Press:
Helsinki).

*Index INDEX

abnormal 96 ball 5
activation of memory 104 basin of attraction 11
additive character 77 Bayes 178
Albeverio 144 behaviour
anomalous 96 depressive 19
arbitrage 212 manic 68
Aref’eva 41 Bergson 93, 107
artificial intelligence 23 blindsight 148
associations 3, 27 Bohm 94
attractor 2 Boolean algebra 28

243
Born 73 Fourier transform 76
Brekke 144 Frampton 41
catecholamine 15 free information wave 78
central neuron 134 frequency domain model 30, 105
cognitive observables 75 Freud 26, 93
cognitive tree 132 function
combinatorial explosion 26 analytic 45
compact group 50 exponential 43
conscious 2 recursive 23
field 86 trigonometric 43
force 86, 90 fuzzy cycle 12
trajectory 91 game of chance 37
consciousness 24, 86 Haar measure 42
collective 87 habituation 17
context Hamiltonian
physical 154 dynamics 93
cognitive 154 equation 95
Copenhagen interpretation 153 hardware 20
creativity 69 Hausdorff 117
crises of creativity 69 Heisenberg 73
cycle 11 helplessness 15
Darwin 102 Hensel lemma 54
Dennett 26 hierarchical
depression 15 chain of neurons 3, 27
derivative 40 coding system 2
Dirac 73 structure 13, 132
discreteness of time 151 Hilbert space probability calculus
Dragovich 41 73
entropy 148 Hiley 94
ergodicity 50 homeostatic 18
excitation 19 homunculus 156
Fatou set 47 ideas 28
financial inferential coherence 24
mass 212 information 1
energy 212 acceleration 95
fundamental constant 220 active 1, 94
kinetic energy 212 field 93
mental force 218 force 95
pilot wave 216 inertia 95
potential energy 212 kinetic energy 95
fixed point 11 mass 95

244
 Newton law 95 Olson 144
passive 1 Ostrovsky theorem 42
time 88 Parisi 41
uncertainty relation 80 Penrose 71
insect 21 period 11
interference of probabilities 154 periodic point 11
isometric cognitive systems 119 physical energy 94
isometry 118 pilot wave 90
Julia set 47 Planck distance 41
Karwowski 144 price acceleration 213
Kolmogorov 178 price momentum 213
lattice 118, 119 psychokinesis 114
law of statistical balance 181 psychological time 150
learning 16 quanta 73
limbic 15 quantum
Markov force 89
chain 65 reductionism 71
dynamics 37 random dynamical systems 57
materialistic axiom 26, 107 random generator 37
measure of mentality 83 relaxation 69
mental rule of reason 13
interference experiment 182 Schrödinger equation 86
phase space 144 Shannon 94
Planck constant 77 Shimony 71
state 133 social system 111
metric 4 software 20
Hausdorff 34 Soros theory of reflexivity 210
pseudo 34 spatial domain model 30
ultra 4, 132 sphere 6
Mises 189 stress 69
motivation 76, 88, 143 string theory 40
observable 76 strong triangle inequality 4
representation 143 synergetic viewpoint 154
neural code 150 temperature of consciousness 147
neural pathway approach 132 thinking processor 2
neuron-activation observable 81 transmitter 15
noise tree 3
psychological 57 Turing machine 23
information 57 ultrametrization 123
non-Archimedean 41 ultra-pseudo-metric 34
noradrenaline 15 unconscious 2

245
Vaxjo interpretation 153 Whitehead 71
Vladimirov operator 76 Witten 41
Volovich 40 Wittgenstein 26

246
 v
q N
uq
N

q

uN−1

vN −1
q q
vN −1
q v2
q q q q q
1 2 N −2 N −1 N

Figure 9: The tree of the space of ideas

wqN
uN q
b
wN −1q b
bqvN
b
b
uN −1q b
b
b
w2` b
b
bvq N −2
u2 q
```
````qv2
q q q q
2 N − 1N
Figure 10: The tree of the space of ‘super-ideas’
cm (0 0)f:0 r:.05 (1 0)f:0 r:.05 (2 0)f:0 r:.05 (3 0)f:0 r:.05 (4 0)f:0 r:.05 (5
0)f:0 r:.05 (6 0)f:0 r:.05 (7 0)f:0 r:.05 (8 0)f:0 r:.05 (9 0)f:0 r:.05
t:V l:.2 w:.2 (0.1 0)(0.9 0) (1.1 0)(1.9 0) (2.1 0)(2.9 0) (3.1 0)(3.9 0) (4.1
0)(4.9 0) (5.1 0)(5.9 0) (6.1 0)(6.9 0) (7.1 0)(7.9 0) (8.1 0)(8.9 0) (9.1 0)(9.9
0)
(2 0.2)h:C v:B S

Figure 11: Centered pathway

cm (0 2)f:0 r:.05 (1 4)f:0 r:.05 (1 5)f:0 r:.05 (1 2)f:0 r:.05 (1 0)f:0 r:.05 (2.5
0)f:0 r:.05 (2 1)f:0 r:.05 (2 2)f:0 r:.1 (3 0)f:0 r:.05 (3 1)f:0 r:.05 (3 3)f:0 r:.05
(3 4)f:0 r:.05 (4 0)f:0 r:.05
(4 2)f:0 r:.05 (5 2)f:0 r:.05 (5 3)f:0 r:.05 (0 1)f:0 r:.05 (1 3)f:0 r:.05 (0 4)f:0
r:.05 t:V l:.2 w:.2 (0 2)(1 2) (1 0)(2 1) (2 1)(2 2) (2.5 0)(2 1) (1 3)(1 2) (1
2)(2 2) (0 1)(1 2) (0 4)(1 4) (1 4)(2 2) (1 5)(1 4) (4 0)(4 -1) (4 0)(5 0) (3
1)(3 0) (3 1)(4 0) (3 1)(4 1) (3 1)(4 2) (2 2)(3 1) (2 2)(3 3) (3 3)(5 2) (3
3)(5 3) (3 3)(3 4) (3 4)(2 4.5) (3 4)(4 4.5) (5 3)(6 2) (5 3)(6 4) (5 2)(6 1)
(.2 .2) (3 4)(3 5) (5 3)(6 2.5) (5 3)(6 3.5) (2 2.2)h:C v:B S

Figure 12: A tree of S-centered pathways: a ‘cognitive tree’

247
 cm (0 0)f:0 r:.05 (1 0)f:0 r:.05
t:V l:.2 w:.2 (0 0)(1 0) (1 0)(1.5 0.5) (1 0)(1.5 -0.5)
(.2 .2) (1.5 0.5)(2 0.5) (1.5 -0.5)(2 -0.5)
(0.2 -0.3)h:C v:B S

Figure 13: cognitive tree — a

cm (0 0)f:0 r:.05 (1 1)f:0 r:.05 (1 -1)f:0 r:.05 (1.5 0.8)f:0 r:.05 (1.5 1.2)f:0
r:.05 (1 0)f:0 r:.05
t:V l:.2 w:.2 (0 0)(1 0) (0 0)(1 -1) (0 0)(1 1) (1 1)(1.5 1.2) (1 1)(1.5 0.8)
(.2 .2) (1 -1)(1.5 -1.2) (1 -1)(1.5 -0.5) (1 0)(1.5 0) (1.5 1.2)(2 1.2) (1.5 0.8)(2
0.8)
(0 -0.3)h:C v:B S

Figure 14: cognitive tree — b

248

View publication stats