Reflection and Its Use

from Science to Meditation

9.3.2004:1019

Henk Barendregt
NIII, Nijmegen University
The Netherlands
Contents

1 Overview 3

2 Reflection and the living cell 12

3 Reflection and Language 15

4 Reflection and Mathematics 28

5 Reflection and Art 43

6 Reflection and Computers 44

7 Reflection and the human mind 51

2
1. Overview

The phenomenon of reflection will be introduced and clarified by examples.

Reflection plays in several ways a fundamental rôle for our existence. Among
other places the phenomenon occurs in life, in language, in computing and in
mathematical reasoning. A fifth place in which reflection occurs is our spiritual
development. In all of these cases the effects of reflection are powerful, even
downright dramatic. We should be aware of these effects and use them in a
responsible way.

Reflection: domain, coding and interaction

Reflection occurs in situations in which there is a domain of objects that all
have active meaning, i.e. specific functions within the right context. Before
turning to the definition itself, let us present the domains relevant for the four
examples. The first domain is the class of proteins. These have indeed spe-
cific functions within a living organism, from bacterium to homo sapiens. The
second domain consists of sentences in natural language. These are intended,
among other things, to make statements, to ask questions, or to influence others.
The third domain consists of (implemented) computable functions. These per-
form computations—sometimes stand alone, sometimes interactively with the
user—so that an output results that usually serves us in one way or another.
The fourth domain consists of mathematical theorems. These express valid
phenomena about numbers, geometric figures or other abstract entities. When
interpreted in the right way, these will enable us to make correct predictions.
Now let us turn to reflection itself. Besides having a domain of meaningful
objects it needs coding and interaction. Coding means that for every object
of the domain there is another object, the (not necessarily unique) code, from
which the original object can be reconstructed exactly. This process of recon-
struction is called decoding. A code C of an object O does not directly possess
the active meaning of O itself. This happens only after decoding. Therefore
the codes are outside the domain, and form the so-called code set. Finally, the
interaction needed for reflection consists of the encounter of the objects and
their codes. Hereby some objects may change the codes, after decoding giving
rise to modified objects. This process of global feedback (in principle on the
whole domain via the codes) is the essence of reflection.
It should be emphasized that just the coding of elements of a domain is not
sufficient for reflection. A music score may code for a symphony, but the two
are on different levels: playing a symphony usually does not alter the written
music1 .
1
However, in aleatory music—the deliberate inclusion of chance elements as part of a
composition—the performance depends on dice that the players throw. In most cases, the
score (the grand plan of the composition) will not alter. But music in which it really does
alter is a slight extension of this idea.

3
Examples of reflection
Having given this definition, four examples of reflection will be presented.

1. Proteins. The first example has as domain the collection of proteins. A

typical protein is shown in the following picture. Its three dimensional structure
can be perceived by looking at the picture with crossed eyes such that the left
and right images overlap.

Figure 1: A schematic display of the protein NGF Homo Sapiens, a nerve growth
factor. Courtesy of the Swiss Institute of Bioinformatics, Peitsch et al. [1995].
ftp://ftp.expasy.org/databases/swiss-3dimage/IMAGES/JPEG/S3D00467.jpg
Each protein is essentially a linear sequence of elements of a set of 20 amino
acids. Because some of these amino acids attract one another, the protein as-
sumes a three dimensional shape that provides its specific chemical meaning.
The sequence of amino-acids for the NGF protein is shown in Fig.2.

Protein: 241 amino acids; molecular weight 26987 Da.

www.ebi.ac.uk/cgi-bin/expasyfetch?X52599

MSMLFYTLIT AFLIGIQAEP HSESNVPAGH TIPQVHWTKL QHSLDTALRR ARSAPAAAIA 60

ARVAGQTRNI TVDPRLFKKR RLRSPRVLFS TQPPREAADT QDLDFEVGGA APFNRTHRSK 120
RSSSHPIFHR GEFSVCDSVS VWVGDKTTAT DIKGKEVMVL GEVNINNSVF KQYFFETKCR 180
DPNPVDSGCR GIDSKHWNSY CTTTHTFVKA LTMDGKQAAW RFIRIDTACV CVLSRKAVRR 240
A 241

Figure 2: Amino acid sequence of NGF Homo Sapiens.

To mention just two possibilities, some proteins may be building blocks for
structures within or between cells, while other ones may be enzymes that enable
life-sustaining reactions. The code-set of the proteins consists of pieces of DNA,

4
a string of elements from a set of four ‘chemical letters’ (nucleotides). Three
such letters uniquely determine a specific amino acid and hence a string of amino
acids is uniquely determined by a sequence of nucleotides, see Alberts et al.
[1993]. A DNA string does not have the meaning that the protein counterparts
have, for one thing because it has not the specific three dimensional folding.
The first advantage of coding is that DNA is much easier to store and du-
plicate than the protein itself. The interaction in this example is caused by a
modifying effect of the proteins upon the DNA. This is also a second advantage
of the protein coding, providing the possibility of change, to be described later.

ACGT-chain: length 1047 base pairs.

www.ebi.ac.uk/cgi-bin/expasyfetch?X52599

agagagcgct gggagccgga ggggagcgca gcgagttttg gccagtggtc gtgcagtcca 60

aggggctgga tggcatgctg gacccaagct cagctcagcg tccggaccca ataacagttt 120
taccaaggga gcagctttct atcctggcca cactgaggtg catagcgtaa tgtccatgtt 180
gttctacact ctgatcacag cttttctgat cggcatacag gcggaaccac actcagagag 240
caatgtccct gcaggacaca ccatccccca agtccactgg actaaacttc agcattccct 300
tgacactgcc cttcgcagag cccgcagcgc cccggcagcg gcgatagctg cacgcgtggc 360
ggggcagacc cgcaacatta ctgtggaccc caggctgttt aaaaagcggc gactccgttc 420
accccgtgtg ctgtttagca cccagcctcc ccgtgaagct gcagacactc aggatctgga 480
cttcgaggtc ggtggtgctg cccccttcaa caggactcac aggagcaagc ggtcatcatc 540
ccatcccatc ttccacaggg gcgaattctc ggtgtgtgac agtgtcagcg tgtgggttgg 600
ggataagacc accgccacag acatcaaggg caaggaggtg atggtgttgg gagaggtgaa 660
cattaacaac agtgtattca aacagtactt ttttgagacc aagtgccggg acccaaatcc 720
cgttgacagc gggtgccggg gcattgactc aaagcactgg aactcatatt gtaccacgac 780
tcacaccttt gtcaaggcgc tgaccatgga tggcaagcag gctgcctggc ggtttatccg 840
gatagatacg gcctgtgtgt gtgtgctcag caggaaggct gtgagaagag cctgacctgc 900
cgacacgctc cctccccctg ccccttctac actctcctgg gcccctccct acctcaacct 960
gtaaattatt ttaaattata aggactgcat ggtaatttat agtttataca gttttaaaga 1020
atcattattt attaaatttt tggaagc 1047

Figure 3: DNA code of NGF Homo Sapiens.

A simple calculation (1047/3 6= 241) shows that not all the letters in the DNA
sequence are used. In fact, some proteins (RNA splicing complex) make a selec-
tion as to what substring should be used in the decoding toward a new protein.

2. Natural language. The domain of the English language is well-known. It

consists of strings of elements of the Roman alphabet extended by the numerals
and punctuation marks. This domain has a mechanism of coding, called quoting
in this context, that is so simple that it may seem superfluous. A string in
English, for example
Maria
has as code the quote of that string, i.e.

‘Maria’.

In Tarski [1933/1995] it is explained that of the following sentences

5
 1. Maria is a nice girl

2. Maria consists of five letters

3. ‘Maria’ is a nice girl

4. ‘Maria’ consists of five letters

the first and last one are meaningful and possibly valid, whereas the second
and third are always incorrect, because a confusion of categories has been made
(Maria consist of cells, not of letters; ‘Maria’ is not a girl, but a proper name).
We see the simple mechanism of coding, and its interaction with ordinary lan-
guage. Again, we see that the codes of the words do not possess the meaning
that the words themselves do.

3. Computable functions. A third example of reflection comes from com-

puting. The first computers made during WW2 were ad hoc machines, each
built for a specific use. Since hardware at that time was a huge investment, it
was recycled by rewiring the parts after each completed job. Based on ideas of
Turing, this procedure was changed. One particular computer was constructed,
the universal machine, and for each particular computing job one had to pro-
vide two inputs: the instructions (the software) and the data that this recipe
acts upon. This has become the standard for all subsequent computers.
3 3
² Input ² Input
¥ || ¥ ¥ || ¥
Output Output
M1 (x) = 2 · x = /6 M2 (x) = x2 = /9

¥ ¥ ¥ ¥

Tabel 3. Two ad hoc machines: M1 for doubling and M2 for squaring a

number.

p1 3
² Program ² Input
¥ || || ¥
Output
UM (p1 , x) = 2 · x = /6

¥ ¥
p2 3
² Program ² Input
¥ || || ¥
Output
UM (p2 , x) = x2 = /9

¥ ¥

6
 Tabel 4. Universal machine UM with programs p1 , p2 simulating M1 , M2
respectively.

So p1 is a code for M1 and p2 for M2 . Since we can consider M1 (p2 ) and M2 (p2 ),
there is interaction: agents acting on a code, in the second case even their own
code.

The domain in this case consists of implemented computable functions, i.e.

machines ready for a specific computing job to be performed. A code for an el-
ement of this domain consists of a program that simulates the job on a universal
machine. The program of a computable function is not yet active, not yet exe-
cutable in computer science terminology. Only after decoding does a program
come into action. Besides coding, interaction is also present. In the universal
machine the program and the data are usually kept strictly separate. But this
is not obligatory. One can make the program and the input data overlap so
that after running for a while on the universal computer, the initial program is
modified.

4. Mathematical theorems. A final example in this section is concerned

with mathematics. A mathematical theorem is usually about numbers or other
abstract entities. Gödel introduced codes for mathematical statements and used
as code-set the collection {0, 1, 2, 3, . . .} of natural numbers, that do not have
any assertive power. As a consequence, one can formulate in mathematics not
only statements about numbers, but via coding also about other such state-
ments. There are even statements that speak about themselves. Again we see
that both the coding and interaction aspects of reflection are present.

The power of reflection

The mentioned examples of reflection all have quite powerful consequences.
We know how dramatically life has transformed our planet. Life essentially
depends on the DNA coding of proteins and the fact that these proteins can
modify DNA. This modification is necessary in order to replicate DNA or to
proof-read it preventing fatal errors.
One particular species, homo sapiens, possesses language. We know its
dramatic effects. Reflection using quoting is an essential element in language
acquisition. It enables a child to ask questions like: “Mother, what is the
meaning of the word ‘curious’ ?”
Reflection in computing has given us the universal machine. Just one design2
with a range of possibilities through software. This has had a multi-trillion US$
impact on the present stage of the industrial revolution of which we cannot yet
see all the consequences.
The effects of reflection in mathematics are less well-known. In this disci-
pline there are statements of which one can see intuitively that they are true,
2
That there are several kinds of computers on the market is a minor detail: this has to do
with speed and user-friendliness.

7
but a formal proof is not immediate. Using reflection, however, proofs using
intuition can be replaced by formal proofs3 , see Howe [1995] and Barendregt
[1997], pp. 21-23. Formal provability is important for the emerging technology
of interactive (human-computer) theorem proving and proof verification. Such
formal and machine-checked proofs are already changing the way hardware is
being constructed4 and in the future probably also on the way one will develop
software. As to the art of mathematics itself, it will bring the technology of
Computer Algebra (dealing exactly
√ with equations between symbolic expres-
sions involving elements like 2 and π) to the level of arbitrary mathematical
statements (involving more complex relations than just equalities between ar-
bitrary mathematical concepts).

The other side of reflection

Anything that is useful and powerful (like fire), can also have a different usage
(such as arson). Similarly the power of reflection in the four given examples
can be used in different ways.
Reflection in the chemistry of life has produced the species, but also it has as
consequence the existence of viruses. Within natural language reflection gives
rise to learning a language, but also to paradoxes5 . The universal computer
has as a consequence that there are unsolvable problems, notably the ones we
are most interested in6 . Reflection within mathematics has as a consequence
that for almost all interesting consistent axiomatic theories, there are state-
ments that cannot be settled (proved or refuted) within that theory (Gödel’s
incompleteness result mentioned above).
We see that reflection may be compared to the forbidden fruit: it is powerful,
but at the same time, it entails dangers and limitations as well. A proper view
of these limitations will make us more modest.

Reflection in spirituality
Insight (vipassana) meditation, which stems from classical Buddhism, concerns
itself with our consciousness. When impressions come to us through our senses,
we obtain a mental representation (e.g. an object in front of us). Now this
mental image may be recollected : this means that we obtain the awareness of
the awareness, also called mindfulness. In order to develop the right mindfulness
3
Often an opposite claim is based on Gödel’s incompleteness result. Given a mathematical
theory T containing at least arithmetic that is consistent (expressed as Con(T )), incomplete-
ness states the following. There is a statement G (equivalent to ‘G is not provable’) within the
language of T that is neither provable nor refutable in T , but nevertheless valid, see Smullyan
[1992]. It is easy to show that G is unprovable if T is consistent, hence by construction G
is true. So we have informally proved that G follows from Con(T ). Our (to some unconven-
tional) view on Gödel’s theorem is based on the following. By reflection one also can show
formally that Con(T )→G. Hence it comes not as a surprise, that G is valid on the basis of
the assumed consistency. This has nothing to do with the specialness of the human mind, in
which we believe but on different grounds, see the section ‘Reflection in spirituality’.
4
Making it much more reliable.
5
Like ‘This sentence is false.’
6
‘Is this computation going to halt or run forever?’ See Yates [1998]

8
it should be applied to all aspects of consciousness. Parts that usually are not
seen as content, but as a coloring of consciousness, become just as important as
the object of meditation. If a leg hurts during meditation, one should be mindful
of it. Moreover, one learns not only to see the pain, but also the feelings and
reactions in connection to that pain. This fine-grained mindfulness will have an
‘intuitive analytic’ effect: our mind becomes decomposed into its constituents
(input, feeling, cognition, conditioning and awareness). Seeing this, we become
less subject to various possible vicious circles in our body-mind system that
often push us into greed, hatred or compulsive thinking.
Because mindfulness brings the components of consciousness to the open in
a disconnected, bare form, they are devoid of their usual meaning. The total
information of ordinary mental states can be reconstructed from mindfulness.
That is why it works like coding with the contents of our consciousness as
domain.
The reflective rôle of mindfulness on our consciousness is quite similar to
that of quoting in ordinary language. As proteins can purify part of our DNA,
the insight into the constituents of consciousness can purify our mind. Mind-
fulness makes visible processes within consciousness, hitherto unseen. After
that, mindfulness serves as a protection by not letting the components of con-
sciousness exercise their usual meaning. Finally, the presence of mindfulness
reorganizes consciousness, giving it a degree of freedom greater than before.
Using mindfulness one may act, even if one does not dare; or, one may abstain
from action, even if one is urged. Then wisdom will result: morality not based
on duty but on virtue. This is the interaction of consciousness and mindfulness.
Therefore, by our definition, one can speak of reflection.
This power of reflection via mindfulness also has another side to it. The
splitting of our consciousness into components causes a vanishing of the usual
view we hold of ourselves and the world. If these phenomena are not accom-
panied in a proper way, they may become disturbing. But during the intensive
meditation retreats the teacher pays proper attention to this. With the right
understanding and reorganization, the meditator obtains a new stable balance,
as soon as one knows and has incorporated the phenomena.
Mental disorders related to stress can cause similar dissociations. Although
the sufferers appear to function normally, to them the world or worse their per-
son does not seem real. This may be viewed as an incomplete and unsystematic
use of mindfulness. Perhaps this explains the enigma of why some of the suf-
ferers become ‘weller than well’, as was observed in Menninger [1963]. These
cured patients might very well have obtained the mental purification that is the
objective of vipassana meditation.

Pure Consciousness
In Hofstadter [1979] the notion of ‘strange loop’ is introduced: ‘Something that
contains a part that becomes a copy of the total when zoomed out. ‘Reflec-
tion’ in this paper is inspired by that notion, but focuses on a special aspect:
zooming out in reflection works via the mechanism of coding. The main the-
sis of Hofstadter is that ‘strange loops’ are at the basis of self-consciousness.

9
I partly agree with this thesis and would like to add that mindfulness serves
as the necessary zooming mechanism in the strange loop of self-consciousness.
On the other hand, the thesis only explains the ‘self’ aspect, the consciousness
part still remains obscure. I disagree with the title of Dennet [1993]: ‘Con-
sciousness explained’. No matter how many levels of cognition and feedback
we place on top of sensory input in a model of the mind, it a priori seems not
able to account for experiences. We always could simulated these processes on
an old-fashioned computer consisting of relays, or even play it as a social game
with cards. It is not that I object to base our consciousness on outer agents
like the card players (we depend on nature in a similar way). It is the claimed
emergence of consciousness as a side effect of the card game that seems absurd.
See Blackmore [2002] for a good survey of theories about consciousness.
Spiritual reflection introduces us to awareness beyond ordinary conscious-
ness, which is without content, but nevertheless conscious. It is called pure
consciousness. This phenomenon may be explained by comparing our person-
ality to the images on a celluloid film, in which we are playing the title role of
our life. Although everything that is familiar to us is depicted on the film, it is
in the dark. We need light to see the film as a movie. It may be the case that
this pure consciousness is the missing explanatory link between the purely neu-
rophysiological activity of our brain and the conscious mind that we (at least
think to) possess. This pure light is believed to transcends the person. The
difference between you and me is in the matter (cf. the celluloid of the film).
That what gives us awareness is said to come from a common source: the pure
consciousness acting as the necessary ‘light’.
To understand where this pure consciousness (our inner light) comes from
we may have to look better into nature (through a new kind of physics, see
e.g. Chalmers [1996] or Stapp [1996]) or better into ourselves (through insight
meditation, see e.g. Goldstein [1983]). Probably we will need to do both.

References
Alberts, B. et al. [1994] The Cell. Garland.
Barendregt, H. [1997] The impact of the lambda calculus,
Bulletin of Symbolic Logic, Volume 3, no 2, 181–215.

Blackmore, S. [2002] Consciousness, an Introduction. Oxford University Press.

Chalmers, D. [1996] The Conscious Mind, Towards a Fundamental Theory,
Oxford University Press.
Dennet, D.C. [1993] Consciousness Explained. Penguin Books.
Harrison, J. [1995] Metatheory and Reflection in Theorem Proving: A Survey and
Critique, available at
http://www.cl.cam.ac.uk/users/jrh/papers/reflect.dvi.gz
Goldstein, J. [1983] The Experience of Insight, Shambhala, Boston, London.
Hofstadter, D. [1979] Gödel Escher Bach. Harvester Press.
Howe, D.J. [1992] Reflecting the semantics of reflected proof, in: Proof Theory,
Aczel et al. (eds.), Cambridge University Press, 229–250.
Menninger, K., M. Mayman and P. Pruyser [1963] The Vital Balance. The Life
Process in Mental Health and Illness. Viking.
Peitsch, M.C., D.R. Stampf, T.N.C. Wells, J.L. Sussman [1995] The Swiss-3DImage

10
 collection and PDB-Browser on the World-Wide Web. Trends in Biochemical
Sciences 20, 82-84. See also: http://www.expasy.org
Smullyan, R. [1992] Gödel’s Incompleteness Theorems. Oxford University Press.
Stapp, H. [1996] The Hard Problem: A Quantum Approach. Journal of Conscious-
ness Studies, 3 (3), 194–210.
Tarski, A. [1933/1995] Introduction to Logic. Dover.
Yates, M. [1998] What computers can’t do. +Plus, issue 5, available at
http://plus.maths.org/issue5/index.html.

Acknowledgement. Reprinted with permission of Templeton Foundation Press from

c 2003. I
the forthcoming Spiritual Information, edited by Charles L. Harper, Jr., °
thank the desk editor Pamela Bond for useful linguistic suggestions.

11
2. Reflection and the living cell
Lecture by Prof.dr. Peter Bloemers. Text From: Mat Ridley, Genome, The Autobiog-
raphy of a Species in 23 Chapters, ed. Fourth Estate Ltd, London 1999, pp. 6-9.

The human body contains approximately 100 trillion (1014 ) cells, most of which
are less than a tenth of a millimeter across. Inside each cell there is a black
blob called a nucleus. Inside the nucleus are two complete sets of the human
genome (except in egg cells and sperm cells, which have one copy each, and red
blood cells, which have none). One set of the genome came from the mother
and one from the father. In principle, each set includes the same 60,000-80,000 7
genes on the same twenty-three chromosomes. In practice, there are often small
and subtle differences between the paternal and maternal versions of each gene,
differences that account for blue eyes or brown, for example. When we breed,
we pass on one complete set, but only after swapping bits of the paternal and
maternal chromosomes in a procedure known as recombination.
Imagine that the genome is a book.

• There are twenty-three chapters, called chromosomes.

• Each chapter contains several thousand stories, called genes.

• Each story is made up of paragraphs, called exons, which are interrupted

by advertisements called introns.

• Each paragraph is made up of words, called codons.

Each word is written in letters called BASES. There are one billion words
in the book, which makes it longer than 5,000 volumes the size of this one8 ,
or as long as 800 Bibles. If I read the genome out to you at the rate of one
word per second for eight hours a day, it would take me a century. If I wrote
out the human genome, one letter per millimeter, my text would be as long as
the River Danube. This is a gigantic document, an immense book, a recipe of
extravagant length, and it all fits inside the microscopic nucleus of a tiny cell
that fits easily upon the head of a pin.
The idea of the genome as a book is not, strictly speaking, even a metaphor.
It is literally true. A book is a piece of digital information, written in linear, one-
dimensional and one-directional form and defined by a code that transliterates
a small alphabet of signs into a large lexicon of meanings through the order
of their groupings. So is a genome. The only complication is that all English
books read from left to right, whereas some parts of the genome read from left
to right, and some from right to left, though never both at the same time.
(Incidentally, you wi1l not find the tired word ’blueprint’ in this book, af-
ter this paragraph, for three reasons. First, only architects and engineers use
blueprints and even they are giving them up in the computer age, whereas we
all use books. Second, blueprints are very bad analogies for genes. Blueprints
7
The present estimate is 30,000 - 40,000. [H.P.J.B.]
8
This text was taken from a book of 344 pages.

12
are two-dimensional maps, not one-dimensional digital codes. Third, blueprints
are too literal for genetics, because each part of a blueprint makes an equivalent
part of the machine or building; each sentence of a recipe book does not make
a different mouthful of cake.)
Whereas English books are written in words of variable length using twenty-
six letters, genomes are written entirely in three-letter words, using only four let-
ters: A, C, G and T (which stand for adenine, cytosine, guanine and thymine).
And instead of being written on flat pages, they are written on long chains of
sugar and phosphate called DNA molecules to which the bases are attached as
side rungs. Each chromosome is one pair of (very) long DNA molecules.
The genome is a very clever book, because in the right conditions it can both
photocopy itself and read itself. The photocopying is known as replication, and
the reading as translation. Replication works because of an ingenious property
of the four bases: A likes to pair with T, and G with C. So a single strand of
DNA can copy itself by assembling a complementary strand with Ts opposite
all the As, As opposite all the Ts, Cs opposite all the Gs and Gs opposite all the
Cs. In fact, the usual state of DNA is the famous double helix of the original
strand and its complementary pair intertwined.
To make a copy of the complementary strand therefore brings back the
original text. So the sequence ACGT becomes TGCA in the copy, which tran-
scribes back to ACGT in the copy of the copy. This enables DNA to replicate
indefinitely, yet still contain the same information.
Translation is a little more complicated. First the text of a gene is tran-
scribed into a copy by the same base-pairing process, but this time the copy is
made not of DNA but of RNA, a very slightly different chemical. RNA, too,
can carry a linear code and it uses the same letters as DNA except that it uses
U, for uracil, in place of T. This RNA copy, called the messenger RNA, is then
edited by the excision of all introns and the splicing together of all exons (see
above).
The messenger is then befriended by a microscopic machine called a ribo-
some, itself made partly of RNA. The ribosome moves along the messenger,
translating each three-letter codon in turn into one letter of a different alpha-
bet, an alphabet of twenty different amino acids, each brought by a different
version of a molecule called transfer RNA. Each amino acid is attached to the
last to form a chain in the same order as the codons. When the whole message
has been translated, the chain of amino acids folds itself up into a distinctive
shape that depends on its sequence. It is now known as a protein.
Almost everything in the body, from hair to hormones, is either made of
proteins or made by them. Every protein is a translated gene. In particular, the
body’s chemical reactions are catalysed by proteins known as enzymes. Even
the processing, photocopying error-correction and assembly of DNA and RNA
molecules themselves - the replication and translation - are done with the help
of proteins. Proteins are also responsible for switching genes on and off, by
physically attaching themselves to promoter and enhancer sequences near the
start of a gene’s text. Different genes are switched on in different parts of the
body.
When genes are replicated, mistakes are sometimes made. A letter (base)

13
is occasionally missed out or the wrong letter inserted. Whole sentences or
paragraphs are sometimes duplicated, omitted or reversed. This is known as
mutation. Many mutations are neither harmful nor beneficial, for instance if
they change one codon to another that has the same amino acid ’meaning’:
there are sixty-four different codons and only twenty amino acids, so many
DNA ’words’ share the same meaning. Human beings accumulate about one
hundred mutations per generation, which may not seem much given that there
are more than a billion codons in the human genome, but in the wrong place
even a single one can be fatal.
A1l rules have exceptions (including this one). Not all human genes are
found on the twenty-three principal chromosomes; a few live inside little blobs
called mitochondria and have probably done so ever since mitochondria were
free-living bacteria. Not all genes are made of DNA: some viruses use RNA
instead. Not all genes are recipes for proteins. Some genes are transcribed into
RNA but not translated into protein; the RNA goes directly to work instead
either as part of a ribosome or as a transfer RNA. Not all reactions are catalysed
by proteins; a few are catalysed by RNA instead. Not every protein comes from
a single gene; some are put together from several recipes. Not all of the sixty-
four three-letter codons specify an amino acid: three signify STOP commands
instead. And finally, not all DNA spells out genes. Most of it is a jumble of
repetitive or random sequences that is rarely or never transcribed: the so-called
junk DNA.
That is all you need to know. The tour of the human genome can begin.

U C A G A Ala
UUU Phe UCU Ser UAU Tyr UGU Cys C Cys
UUC Phe UCC Ser UAC Tyr UGC Cys D Asp
U
UUA Leu UCA Ser UAA stop UGA stop E Glu
UUG Leu UCG Ser UAG stop UGG Trp G Gly
CUU Leu CCU Pro CAU His CGU Arg F Phe
CUC Leu CCC Pro CAC His CGC Arg H His
C I Ile
CUA Leu CCA Pro CAA Gln CGA Arg
CUG Leu CCG Pro CAG Gln CGG Arg K Lys
AUU Ile ACU Thr AAU Asn AGU Ser L Leu
AUC Ile ACC Thr AAC Asn AGC Ser M Met
A N Asn
AUA Ile ACA Thr AAA Lys AGA Arg
AUG Met ACG Thr AAG Lys AGG Arg P Pro
GUU Val GCU Ala GAU Asp GGU Gly Q Gln
GUC Val GCC Ala GAC Asp GGC Gly R Arg
G S Ser
GUA Val GCA Ala GAA Glu GGA Gly
GUG Val GCG Ala GAG Glu GGG Gly T Thr
V Val
W Trp
Y Tyr

Figure 4: The ‘universal’ genetic code and the naming convention for
aminoacids. Three codons (UAA, UAG and UGA) code for the end of a protein
(‘stop’).

14
3. Reflection and Language

Describing and generating languages is an important subject of study.

Formal languages are precisely defined via logical rules. These languages are
introduced for special purposes. For example the programming languages de-
scribe algorithms, i.e. calculation recipes, used to make computers do all kinds
of (hopefully) useful things. Other formal languages can express properties of
software, the so-called specification languages, or properties part of some math-
ematical theory. Finally, some formal languages are used in order to express
proves of properties, for example the proof that a certain program does what
you would like it to do9
Natural languages occur in many places on earth and are used by people
to communicate. Part of linguistics uses ideas of formal languages in order to
approach better and better the natural languages. The hope cherished by some
is to be able to come up with a formal description of a large part, if not the
total, of the natural languages. We will discuss mainly formal languages, giving
only a hint how this study is useful for natural ones.
A language is a collection of words. A word is a string of symbols taken
from a predefined alphabet. A typical questions are
• Does word w belong to language L?

• Are the languages L and L0 equal?

Words over an alphabet

3.1. Definition. (i) An alphabet Σ is a set of symbols. Often this set is finite.
(ii) Given an alphabet Σ, a word over Σ is a finite sequence w = s1 . . . sn of
elements si ∈Σ. It is allowed that n = 0 in which case w = ² the empty word.
(iii) We explain the notion of an abstract syntax by redefining the collection
of words over Σ as follows:

word := ² | word s ,
where s∈Σ.
(iv) Σ∗ is the collection of all words over Σ.

3.2. Example. (i) Let Σ1 = {0, 1}. Then

1101001∈Σ∗1 .

(ii) Let Σ2 = {a, b}. Then

abba ∈ Σ∗2 .
/ Σ∗2 .
abracadabra ∈

(iii) abracadabra∈Σ∗3 , with Σ3 = {a, b, c, d, r}.

9
If software is informally and formally specified, tested and proven correct, i.e. that it
satisfies the specification it obtains five stars. The informal specification and tests serve to
convince the reader that the requirements are correctly stated. There is very little five star
software.

15
 (iv) ²∈Σ∗ for all Σ.
(v) Let Σ4 = {A, C, D, E, F, G, H, I, K, L, M, N, P, Q, R, S, T, V, W, Y}. Then the fol-
lowing is a word in Σ∗4 .
MSMLFYTLITAFLIGIQAEPHSESNVPAGHTIPQVHWTKLQHSLDTALRRARSAPAAAIA
ARVAGQTRNITVDPRLFKKRRLRSPRVLFSTQPPREAADTQDLDFEVGGAAPFNRTHRSK
RSSSHPIFHRGEFSVCDSVSVWVGDKTTATDIKGKEVMVLGEVNINNSVFKQYFFETKCR
DPNPVDSGCRGIDSKHWNSYCTTTHTFVKALTMDGKQAAWRFIRIDTACVCVLSRKAVRRA

We have encountered it in Fig. 2 of section 1.

(vi) Let Σ5 = {a, c, g, t}. Then an element of Σ∗5 is given in Fig. 3.
(vii) Let Σ6 = {a, c, g, u}. Then Σ6 is “isomorphic to” Σ5 .

Operations on words
3.3. Definition. (i) If a∈Σ and w∈Σ∗ , then a.w is defined ‘by induction on
w’.

a.² = a
a.(us) = (a.u)s

(ii) If w, v∈Σ∗ , then their concatenation

w++v

in Σ∗ is defined by induction on v:

w++² = w
w++us = (w++u)s.

We write wv ≡ w++v as abbreviation.

(iii) Let w∈Σ∗ . Then w∨ is w “read backward” and is formally defined by

²∨ = ²;
(wa)∨ = a(w∨ ).

For example (abba)∨ = abba, and (abb)∨ = bba.

Languages
3.4. Definition. Let Σ be an alphabet. A language over Σ is just a subset
L ⊆ Σ∗ (defined in one way or another).

3.5. Example. Let Σ = {a, b}. Define the following languages over Σ.
(i) L1 = {w | w starts with a and ends with b}. Then

ab, abbab∈L1 , but ², abba, bab ∈

/ L1 .

(ii) L2 = {w | abba is part of w}. Then

abba, abbab∈L2 , but ², ab, bab ∈

/ L2 .

16
 (iii) L3 = {w | aa is not part of w}. Then

abba, abbab∈L3 , but ², aa, babaa ∈

/ L3 .

(iv) L4 = {², ab, aabb, aaabbb, . . . , an bn , . . .}

= {an bn | n ≥ 0}.
Then ², aaaabbbb∈L4 but aabbb, bbaa ∈
/ L4 .
(v) L5 = {w | w is a palindrome, i.e. w = w ∨ }. For example abba∈L5 , but
abb ∈
/ L5 .

Operations on languages
3.6. Definition. Let L, L1 , L2 be languages over Σ. We define

L1 L2 = {w1 w2 ∈Σ∗ | w1 ∈L1 & w2 ∈L2 }.

L1 ∪ L2 = {w∈Σ∗ | w∈L1 or w∈L2 }.
L∗ = {w1 w2 . . . wn | n ≥ 0 & w1 , . . . , wn ∈L}.
L+ = {w1 w2 . . . wn | n > 0 & w1 , . . . , wn ∈L}.

Some concrete languages

3.7. Definition. (Hofstadter) Let ΣH = {M, I, U}. Define the language LH
over ΣH by the following grammar, where x, y∈Σ∗H .

axiom MI
rules xI ⇒ xIU
Mx ⇒ Mxx
xIIIy ⇒ xUy
xU Uy ⇒ xy

This means that by definition MI∈LH ;

if xI∈LH , then also xIU∈LH ,
if Mx∈LH , then also Mxx,
if xIIIy∈LH , then also xUy,
if xU Uy∈LH , then also xy.

3.8. Example. (i) MI, MII, MU, MUU, IMMIU, . . .∈Σ∗H

(ii) MI, MIU, MIUIU, MII, MIIII, . . .∈LH .

3.9. Exercise. (i) Show that MUI∈LH

(ii) Show that IMUI ∈
/ LH

3.10. Problem (Hofstadter’s MU puzzle10 ). MU∈LH ?

How would you solve this?

3.11. Definition. (i) Σ2 = {p, q, −}

10
See Hofstadter [1979].

17
 (ii) The language L2 over Σ2 is defined as follows (x, y, z∈Σ∗2 ).

axioma’s xpqx if x consists only of −s

rule xpyqz ⇒ xpy−qz−

3.12. Exercise. Which words belong to L2 ? Motivate your answers.

1. −−p−−p−−q−−−−

2. −−p−−q−−q−−−−

3. −−p−−q−−−−

4. −−p−−q−−−−−−

3.13. Exercise. Let Σ3 = {a, b, c}. Define L3 by

axiom ab
rule xyb ⇒ yybx

The following should be answered by ‘yes’ or ‘no’, plus a complete motivation

why this is the right answer.
(i) Do we have ba∈L3 ?
(ii) Do we have bb∈L3 ?

Even languages over a single letter alphabet are interesting.

3.14. Definition. Laat Σ4 = {a}.

(i) Define L41 as follows.

axiom a
rule w ⇒ waa

Then L41 = {an | n is an odd number}. Here one has a0 = λ and an+1 = an a.
In other words an = a
| .{z
. . a}.
n times
(ii) L42 = {ap | p is a prime number}.
(iii) Define L43 as follows.

axiom a
rule w ⇒ ww
wwwa ⇒ w

How can one decide whether Σ4 is in L41 ? The question ‘w∈L42 ?0 is more
difficult. The difficulty is partly due to the specification of L42 . Language L43
has an easy grammar, but a difficult decision problem. For example it requires
several steps to show that aaa∈L43 .
Challenge. Do we have L43 = {an | n ≥ 1}? The first person who sends via
email the proof or refutation of this to <[email protected]> will obtain 100 = C.

18
Closing time 1.05.2004.

Open problem. (Collatz’ conjecture) Define L44 as follows.

axiom a
rule w ⇒ ww
wwwaa ⇒ wwa

Prove or refute Collatz’ conjecture

L44 = {an | n ≥ 1}.

The first correct solution by email before 1.05.2004 earns 150 =

C . Is there a
relation bewteen L43 and L44 ?

3.15. Exercise. (i) Show that ² ∈

/ L44 .
(ii) Show that aaa∈L44 .

Regular languages
Some of the languages of Example 3.5 have a convenient notation.

3.16. Example. Let Σ = {a, b}. Then

(i) L1 is denoted by a(a ∪ b)∗ b.
(ii) L2 is denoted by (a ∪ b)∗ abba(a ∪ b)∗ .

3.17. Definition. Let Σ be an alphabet.

(i) The regular expressions over Σ are defined by the following grammar

re := ∅ | ² | s | (re.re) | (re ∪ re) | re∗ .

here s is an element of Σ. A more consice version of this grammar is said to be

an abstract syntax :

re := ∅ | ² | s | re.re | re ∪ re | re∗ .

(ii) Given a regular expression e we define a language L(e) over Σ as follows.

L(∅) = ∅;
L(²) = {²};
L(s) = {s};
L(e1 e2 ) = L(e1 )L(e2 );
L(e1 ∪ e2 ) = L(e1 ) ∪ L(e2 );
L(e∗ ) = L(e)∗

(iii) A language L over Σ is called regular if L = L(e) for some regular

expression e.

19
Note that L+ = L.L∗ so that it one may make use of + in the formation of
regular languages. Without a proof we state the following.

3.18. Proposition. (i) L3 of Example 3.5 is regular:

L3 = L((b ∪ ab)∗ (a ∪ ²)).

(ii) L4 = {an bn | n ≥ 0} is not regular.

There is a totally different definition of the class of regular languages, namely

those that are “accepted by a finite automaton”. The definition is not compli-
cated, but beyond the scope of these lectures.

Contextfree languages
There is another way to introduce languages. We start with an intuitive ex-
ample. Consider the following production system (grammar) over the alphabet
Σ = {a, b}.
S → ² | aSb

This is nothing more or less than the grammar

exp := ² | a exp b

The S stands for start. With this auxiliary symbol we start. Then we follow the
arrow. There are two possibilities: ² and aSb. Since the first does not contain
the auxiliary symbol any longer, we say that we have reached a terminal state
and therefore the word ² has been produced. The second possibility yields aSb,
containing again the ‘non-terminal’ symbol S. Therefore this production has
not yet terminated. Continuing we obtain

ab = a²b and aaSbb.

And then
aabb and aaaSbbb.
Etcetera. Therefore this grammar generates the language

L5 = {², ab, aabb, aaabbb, a4 b4 , . . . , an bn , . . .},

also written as
L5 = {an bn | n ≥ 0}.
The productions can be depicted as follows.

S → ²;
S → aSb → ab;
S → aSb → aaSbb → aabb;
S → aSb → aaSbb → aaaSbbb → aaabbb.

20
 L5 as defined above is called a contextfree language and its grammar a
contextfree grammar
A variant is
S → ab | aSb

generating
L05 = {an bn | n > 0}.

3.19. Definition. A contextfree grammar consists of the following.

(i) An alphabet Σ.
(ii) A set V of auxiliary11 symbols. Among them S, the start symbol.
(iii) A finite collection production rules of the form

X→w

where X is an auxiliary symbol and w a word consisting of letters from the

alphabet and the auxiliary symbols together; otherwise said w∈(Σ ∪ V )∗ , where
∪ denotes the union of two sets.
(iv) If there are two production rules with the same auxiliary symbol as its
left hand side, for example X → w1 and X → w2 , then we notate this in the
grammar as
X → w 1 | w2 .

For the auxiliary symbols we use upper case letters like S, A, B. For the
elements of the alphabet we use lower case letters like a, b, c etcetera.

3.20. Example. (i) L5 , L05 above are contextfree languages. Indeed, the con-
textfree grammars are given.
(ii) L41 = {an | n odd} over Σ = {a} is context-free. Take V = {S} and as
production-rules
S → aaS | a

(iii) Define L7 over Σ = {a, b} using V = {S, A, B} and the production-rules

S → AB
A → Aa | ²
B → Bb | ²

Then L7 = L(a∗ b∗ ), i.e. all string a’s followed by a string b’s.

Note that the auxiliary symbols can be determined from the production- rules.

11
These are also called non-terminal symbols.

21
3.21. Exercise. Given is the grammar

S → aSb | A | ²
A → aAbb | abb

This time V = {S, A} and Σ = {a, b}.

Can one produce abb and aab?
What is the collection of words in the generated language?

3.22. Exercise. Produce the language of 3.21 with axioms and rules as in 3.7.

The name ‘Context-free grammars’ refers to the fact that the left-hand side
of the production rules consist of single auxiliary symbols. (For example the
rule Sa → Sab is not allowed.) One never needs to look at the context in which
the auxiliary symbol is standing.
An important restriction on the context-free grammars consists of the right-
linear grammars.

3.23. Definition. A right-linear grammar is a context-free grammar such that

in every production rule
X→w
one has that w is of one of the following forms
(i) w = ²
(ii) w = vY with v∈Σ∗ and Y an auxiliary symbol.

That is to say, in a right-linear grammar auxiliary symbols on the right of a

rule only stand at the end and only as a single occurrence.

3.24. Example. (i) In Example 3.20 only L41 is a right-linear grammar.

(ii) Sometimes it is possible to transform a context-free grammar in an equiv-
alent right-linear one. The following right-linear grammar (over Σ = {a, b}) also
produces L7 example 3.20 (iii).

S → aS | B
B → bB | ²

Without a proof we state the following.

3.25. Theorem. Let L be a language over Σ. Then

L is regular ⇔ L has a right-linear grammar.

Hence every regular language is context-free.

22
3.26. Exercise. (i) Consider the context-free grammar over {a, b, c}

S → A|B
A → abS | ²
B → bcS | ²

Which of the following words belong to the corresponding language L 8 ?

abab, bcabbc, abba.
(ii) Show that L8 is regular by giving the right regular expression.
(iii) Show that L8 has a right-linear grammar.
3.27. Exercise. Let Σ = {a, b}.
(i) Show that L41 = {an | n is odd}, see 3.20, is regular. Do this both
by providing a regular expression e such that L41 = L(e) and by providing a
right-linear grammar for L41 .
(ii) Describe the regular language L(a(ab∗ )∗ ) by a context-free grammar.
(iii) Let L9 consists of words of the form
aba . . . aba
(i.e. a’s b’s alternating, starting with an a and ending with one; a single a is
also allowed). Show in two ways that L9 is regular.
3.28. Exercise. Let Σ = {a, b, c}. Show that
L = {w∈Σ∗ | w is a palindrome}
is context-free.
3.29. Example. We now give a grammar for a small part of English. Not all
sentences produced by this grammar make sense, though!

S = hsentencei → hnoun − phraseihverb − phrasei.

hsentencei → hnoun − phraseihverb − phraseihobject − phrasei.
hnoun − phrasei → hnamei | harticleihnouni
hnamei → John | Jill
hnouni → bicycle | mango
harticlei → a | the
hverb − phrasei → hverbi | hadverbihverbi
hverbi → eats | rides
hadverbi → slowly | f requently
hadjective − listi → hadjectiveihadjective − listi | ²
hadjectivei → big | juicy | yellow
hobject − phrasei → hadjective − listihnamei
hobject − phrasei → harticleihadjective − listihnouni

23
3.30. Exercise. (i) Show how to produce the sentence

Jill frequently eats a big juicy yellow mango.

(ii) Is the generated language context-free?

(iii) Is this grammar right-linear?
(iv) Produce some sentences of your own.
(v) What is the alphabet Σ?
(vi) What are the auxiliary symbols?

Other classes of languages: the Chomsky hierarchy

3.31. Definition. Let Σ be an alphabet.
(i) A context-sensitive language over Σ is introduced like a context-free lan-
guage by production rules of the form

uXv → uwv,

where u, v, w∈Σ∗ and w 6= ². Here X is an auxiliary symbol. The difference

between these languages and the context-free ones is that now the production
of
X→w
only is allowed within the context

u . . . v.

(ii) The enumerable languages over Σ are also introduced by similar gram-
mars, but now the production rules are of the form

uXv → uwv,

where w = ² is allowed.
(iii) A language L over Σ is called computable if and only if both L and L
are enumerable. Here L is the complement of L:

L = {w∈Σ∗ | w ∈
/ L}.

A typical context-sensitive language is

{an bn cn | n ≥ 0}.

A typical computable language is

{ap | p is a prime number}.

A typical enumerable language is L44 .

The following families of languages are strictly increasing:

1. The regular languages;

2. The context-free languages;

24
 3. The context-sensitive languages;

4. The computable languages;

5. The enumerable languages.

Let us abbreviate these classes of languages as RL,CFL, CSL, CL, EL, respec-
tively. Then we have the proper inclusions can be depicted in the following
diagram.

RL
CFL
CSL
CL
EL

Figure 5: The Chomsky hierarchy

Chomsky discusses the power of these definition mechanisms for the gener-
ation of natural languages. He argues that a natural language is too complex to
be described by a context-free grammar. Moreover, Chomsky argues that the
computable and enumerable languages are too complex to be able to learn by
three year old children. The open problem of linguistics is whether a natural
language can be described as a context-sensitive language.

Reflection over the classes of languages

There is a uniform way to describe the regular languages. By definition a
language L is regular if and only if there is a regular expression e such that
L = L(e). This set of regular expressions is itself not a regular language.
Reflection over the regular languages pushes us outside this class.

3.32. Definition. (i) A universal notation system for the regular languages
over an alphabet Σ consists of a language Lu over an alphabet Σu and a decoding
d : Lu →{L | L is regular}, such that for every regular language L there is at
least one code c such that d(c) = L.
(ii) Such a universal coding system is called regular if the language

{cv | v∈d(c)}

over the alphabet Σ ∪ Σu is regular.

3.33. Proposition. (V. Capretta) There is no regular universal notation sys-

tem for the regular languages.

25
We will not give a proof, as it requires some knowledge about the regular
languages.
A similar negative result is probably also valid for the context-free and
context-sensitive languages. We know that this negative result is the case for
the computable languages. But for the enumerable languages there does exists
a notation system that itself is enumerable.

3.34. Definition. (i) A universal notation system for the enumerable lan-
guages over an alphabet Σ consists of a language Lu over an alphabet Σu
and a decoding d : Lu →{L | L is enumerable}, such that for every enumerable
language L there is at least one code c such that d(c) = L.
(ii) Such a universal coding system is called enumerable if the language

{cv | v∈d(c)}

over the alphabet Σ ∪ Σu is enumerable.

3.35. Proposition. There is an enumerable universal notation system for the

enumerable languages.

Proof. (Sketch) The reason is that the enumerable languages are those lan-
guages that are accepted by a Turing machine. Turing machines take as input
a string w and start a computation, that can halts or not. Now L is enumerable
if and only if there is a Turing machine ML such that

w∈L ⇔ ML (w) halts.

There is an universal Turing machine Mu , see section 6. This means that for
every Turing machine M there is a code cM such that

M (w) = Mu (cM w).

Define f (c) = {w | M (cw) halts}. Then given an enumerable language L one

has

w∈L ⇔ ML (w) halts

⇔ M (cML w) halts
⇔ w∈d(cML ),

hence
L = d(cL ).
Therefore Lu = {cM | M a Turing machine} with decoding d is a universal
notation mechanism for the enumerable languages. Moreover, the notation
system is itself enumerable:

{cw | w∈d(c)} = {cw | Mu (cw) halts},

which is the languages accepted by LMu and hence enumerable.

26
 We end this section by observing that the reflection of the enumerable lan-
guages is a different from the one that is present in the natural language like
English, see section 1. The first one has as domain the collection of enumer-
able languages; the second one has as domain the collection of strings within a
language.

References
Chomsky, N. [1956]. Three models of the description of language, IRE Trans-
actions on Information Theory, 2, 3, pp. 113-124.

Hofstadter, D. [1979] Gödel Escher Bach. Harvester Press.

Sipser, M. [1996] Introduction to the theory of computation, PWS Publishers,

Boston.

27
4. Reflection and Mathematics

The ongoing creation of mathematics, that started 5 or 6 millennia ago and

is still continuing at present, may be described as follows. By looking around
and abstracting from the nature of objects and the size of shapes homo sapiens
created the subjects of arithmetic and geometry. Higher mathematics later
arose as a tower of theories above these two, in order to solve questions at
the basis. It turned out that these more advanced theories often are able to
model part of reality and have applications. By virtue of the quantitative, and
even more qualitative, expressive force of mathematics, every science needs this
discipline. This is the case in order to formulate statements, but also to come
to correct conclusions.
Mathematics consists of the study of patterns. In fact, it makes invisible
things visible (K. Devlin). Think of magnetic force that is invisible, being
visualized by the mathematical laws how moving electric charge and magnetism
mutually influence each other.

Maxwell’s Equations
~ = 0
∇·B
~ + ∂ B/∂t
∇×E ~ = 0
~ = ρ
∇·D
~ − ∂ D/∂t
∇×H ~ = J~

Moreover, using this visualization one obtains some kind of control over the
phenomenon. These mathematical laws are essential in order to build monitors
for televisions and computers, as one needs to know how electron trajectories
are being bent in magnetic fields.

The nature of mathematics

The Greek philosopher Aristotle (384-322 BC) made several fundamental con-
tributions to the foundations of mathematics that are still relevant today. From
him we have inherited the idea of the axiomatic method 12 , not just for math-
ematics, but for all sciences. A science consists of statements about concepts.
Concepts have to be defined from simpler concepts. In order to prevent an infi-
nite regression, this process starts from the primitive concepts, that do not get
a definition. Statements have to be proved from statements obtained before.
Again one has to start somewhere; this time the primitive statements are called
axioms. A statement derived from the axioms by pure reason is called a theorem
in that axiomatic system. In mathematics one starts from arbitrary primitive
notions and axioms, while in science from empirical observations, possibly using
(in addition to pure reason) the principle of induction (generalization).
12
In Aristotle [350 B.C.], Posterior Analytics.

28
 Just a couple of decades after Aristotle and the axiomatic method, Euclid
came with his compilation of existing geometry in this form in his Elements 13 ,
see Euclid [300 BC] and was very influential as an example of the use of the
axiomatic method. Commentators of Euclid stated that the primitive notions
are so clear that they did not need definitions; similarly it was said that the
axioms are so true that they did not need a proof. This, of course, is somewhat
unsatisfactory.
A couple of millennia later Hilbert (1862-1943) changed this view. For him
it did not matter what exactly is the essence of the primitive notions such as
point and line, as long as they satisfy the axioms: “The axioms form an implicit
definition of the primitive concepts”.
In the light of the axiomatic method the act of creating mathematics can
be seen as follows.

Axioms

Reasoning

?
' $

Mathematics

& %

The boundary around “Mathematics” is rounded off, in order to indicate that

at this moment we have to view it as an open ended collection of theorems.
The is because at this stage the notion of “reasoning” is pretty vague.
13
As was already observed in antiquity the theorems in the Elements were not always proved
from the axioms by logic alone. Sometimes his arguments required extra assumptions. The
axiomatization of Hilbert [1899] corrected the subtle flaws in Euclid.

29
Some mathematical phenomena
4.1. Example. Consider the sequence

1, 4, 9, 16, 25, . . .

What is the next element in this? It is 36 and after that follows 49. We have
the sequence of squares. Now we write the difference between two consecutive
elements.

1 4 9 16 25 36 ...

3 5 7 9 11 ...
We see that the odd numbers appear. Some questions come to our mind.

• Will this be going on forever?

• Why do we miss the odd number 1 at the beginning?

Let us start with the second question. We want things to become beautiful. We
should have started counting with 0 (as do all mathematicians and Montessori
children).

0 1 4 9 16 25 36 ...

1 3 5 7 9 11 ...

Much nicer indeed! The first question is equivalent with saying that the second
difference sequence is constantly 2:

0 1 4 9 16 25 36 ...

1 3 5 7 9 11 ...

2 2 2 2 2 ...

More formally we have the following proposition.

Proposition. Let an = n2 . Define bn = an+1 − an , and cn = bn+1 − bn . Then

for all n one has cn = 2.

Proof. bn = an+1 − an
= (n + 1)2 − n2
= n2 + 2n + 1 − n2
= 2n + 1
indeed the odd numbers.

A more elaborate result is the following.

Definition. Let an be a sequence. Define Da as the sequence bn = an+1 − an .

30
Proposition. Let a3n = n3 . Then DDDa3 = 6 for all n.

0 1 8 27 64 125 ...

1 7 19 37 61 ...

6 12 18 24 ...

6 6 6 ...

Still better is the following. Define k! = 1.2.3. . . . .k

Proposition. Let akn = nk . Then D k ak = k! for all n.

Some results and open problems

Definition. A natural number p is called prime (number) iff (if and only if)
p > 1 and every divisor of p is either 1 or p itself. For example

2, 3, 5, 7, 11, 13

are prime numbers, but

5, 9, 15 and 21
are not.

Proposition. (Euclid) There are infinitely many prime numbers. This can be
stated without mentioning the concept of infinity:

For every number n there exists a greater prime number p.

In symbols
∀n∃p [p > n & p is prime].
Proof. Given n. Consider k = n! + 1, where n! = 1.2.3. . . . .n.
Let p be a prime that divides k.
For this number p we have p > n: otherwise p ≤ n;
but then p divides n!, so p cannot divide k = n! + 1,
contradicting the choice of p.

More difficult to prove is the following.

Theorem. (Chebychev) ∀n∃p [n < p < 2n & p is prime].

Open problems. (i) Is every even number > 2 the sum of two primes? E.g.

31
one has
4 = 2+2
6 = 3+3
8 = 3+5
10 = 3+7 = 5+5
12 = 5+7
14 = 7+7 = 3 + 11
16 = 3 + 13 = 5 + 11
18 = 7 + 11 = 5 + 13
20 = 3 + 17 = 7 + 13
22 = 3 + 19 = 5 + 17 = 11 + 11
(ii) Are there infinitely many “prime twins”? A prime twin consists of a
pair of numbers p, p + 2 such that p and p + 2 are both prime. E.g. the first
prime twins are

(3, 5), (5, 7), (11, 13), (17, 19), (29, 31), . . . .

Peano Arithmetic
Arithmetic is the theory about the natural numbers. It contains propositions
like the ones above. Many such proposition scan be proved from the so-called
axioms of Peano. It axiomatizes the properties of the set of natural numbers

0, 1, 2, 3, . . .

Following Hilbert, these will be defined via the so-called Peano axioms. The
number 0 is taken as primitive. Rather than taking the other numbers as
primitive, the successor S, that makes from one number the next one, is taken
as primitive.

1. 0 is a natural number.

2. If n is a natural number, also Sn is a natural number.

3. Sn = Sm→n = m.

4. Sn 6= 0, for all natural numbers n.

5. Let P be a property of natural numbers. Suppose that

P (0)
P (n)→P (S(n)) for all natural numbers n.
Then P (n) for all natural numbers n.

Axiom 5 was first formulated by Blaise Pascal and is called the principle of
mathematical induction. To understand the principle, think of the natural num-
bers as standing in a long row.

0, 1, 2, 3, . . .

32
If a number has property P , then we mark it. We start with marking the first
number, because we know that P (0) holds.

0∨ , 1, 2, 3, . . .

But we know also that P (n)→P (n+1). This means that if a number is marked,
so is its successor. Therefore

0∨ , 1∨ , 2, 3, . . .

and
0∨ , 1∨ , 2∨ , 3, . . .
0∨ , 1∨ , 2∨ , 3∨ , . . .
...
and we understand that all numbers get marked. It is the principle of dominos
standing one next to the other. If the first falls, then it will cause the next to
fall, and then the next, etcetera. Eventually all will fall. This is for the natural
numbers the way how we get within a finite amount of time some grip over all
of them.
Note that there is no notation for the numbers 1, 2, 3, . . .. These numbers
do not occur in the language of PA (but live in Plato’s paradise). Numbers can
be represented as linguistic entities called numerals as follows.

0 = 0;
n+1 = S( n ).

For example 3 = S(S(S(0))).

Addition and multiplication

These operations can be specified as follows.

a+0 = a
a + S(b) = S(a + b).
a·0 = 0
a · S(b) = (a · b) + a.

Proposition. ∀a, b, c (a + b) + c = a + (b + c).

Proof. Given a, b we have to show ∀c P (c), where P (c) := (a + b) + c =
a + (b + c). We do this by mathematical induction.
Case c = 0. Then P (c) states (a + b) + 0 = a + (b + 0). This holds:

(a + b) + 0 = a + b,
= a + (b + 0).

33
 Induction step. Suppose P (c) holds, i.e. (a + b) + c = a + (b + c). We call
this the induction hypothesis. We must show P (S(c)) i.e. (a + b) + S(c) =
a + (b + S(c)). Indeed,

(a + b) + S(c) = S((a + b) + c)
= S(a + (b + c)), by the induction hypothesis,
= a + S(b + c),
= a + (b + S(c)).

Logic
It was again Aristotle who started the quest for logic, i.e. the laws by which
scientific reasoning is possible14 . Aristotle came up with some syllogisms (valid
reasoning step based on syntactical form) like

No A is a B
.
No B is a A
Aristotle explains this by the particular case
No horse is a man
.
No man is a horse
Another of his syllogisms is
No A is a C All B are C
.
No A is a B
Take e.g. men, birds and swans for A, B and C respectively. Aristotle also
makes a distinction between such syllogisms and so called imperfect syllogisms,
that require more steps (nowadays these are called admissible rules). The idea
of specifying formal rules sufficient for scientific reasoning was quite daring and
remarkable at the time. Nevertheless, from a modern perspective the syllogisms
of Aristotle have the following shortcomings. 1. Only unary predicates are
used (monadic logic). 2. Only composed statements involving → and ∀ are
considered (so &, ∨, ¬ and ∃ are missing). 3. The syllogisms are not sufficient
to cover all intuitively correct steps.
In commentators of Aristotle one often finds the following example.
All men are mortal Socrates is a man
. (1)
Socrates is mortal
Such ‘syllogisms’ are not to be found in Aristotle, but became part of the
traditional logical teaching. They have an extra disadvantage, as they seem to
imply that they do need to lead from true sentences to true sentences. This is
14
In Aristotle [350 B.C.], Prior Analysis. One may wonder whether his teacher Plato (427-
347 BC) was in favor of this quest (because we already know how to reason correctly).

34
not the case. Syllogism only need to be truth preserving, even if that truth is
hypothetical. So a more didactic (and more optimistic) version of (1) is

All sentient beings are happy Socrates is a sentient being

. (2)
Socrates is happy
This example is more didactic, because one of the premises is not true, while
the rule is still valid. Aristotle was actually well aware of this hypothetical
reasoning.
It was more than 2300 years later that Frege (1848-1925) completed in 1879
the quest for logic and formulated (first-order) predicate logic. He was helped in
this by Vieta (1540-1603), who introduced variables to denote arbitrary quan-
tities, Leibniz (1646-1716), who axiomitized equality, Boole (1815-1864), who
treated connectives like ‘and’, ‘or’, ‘implies’ and ‘not’, and Peirce (1839-1914),
who studied the quantifiers ‘for all’ and ‘there exists’. That Frege’s logic was
sufficient for the development of mathematics from the axioms was proved in
1922 by Skolem (1887-1963) and independently in 1930 by Gödel (1906-1978).
This result is called the completeness theorem for first order logic.
The axiom system for logic has rules that determine the meaning for all the
logical connectives: ¬, &, ∨, ∀, ∃, that stand for ‘not’, ‘and’, ‘or’, ‘for all’ and
‘exits’ respectively. For example the rules concerning & are
A B A&B A&B
A&B A B
The first rule has to be read as: if statements A and B are given, then one can
deduce A&B. The second rule states that if A&B is given, then on can deduce
A. Other rules are slightly more complex and are beyond these lectures. For
an introduction, see van Dalen [1994].

Definition. The language of Peano arithmetic by a context-free abstract

grammar. We ned the syntactical categories of variables, terms and formulas.

var := x | var 0
term := var | 0 | S term | term + term15 | term · term
f orm := term = term | ¬f orm | f rom ∨ f orm | f orm&f rom |
f orm→f orm | ∀ var f orm | ∃ varf orm

15
Terms like x + y could be avoided by introducing a predicate A(x, y, z) with the intended
meaning
A(x, y, z) ↔ x + y = z.
This A then should satisfy

A+ (x, 0, z) ↔ x = z;
A+ (x, y, z) ↔ A+ (x, S(y), S(z)).

Something similar can be done for multiplication via a predicate A× . But this is not so
convenient. A familiar equation like

(x + y) · z = x · z + y · z

35
We have for example
var = {x, x0 , x00 , x000 , . . .}.
One uses x, y, z, . . . , x1 , y1 , z1 , . . . , a, b, c, . . . to denote arbitrary variables.
One uses s, t, u, . . . , s1 , t1 , u1 , . . . to denote arbitrary terms.
One uses P, Q, R, . . . , A, B, C, . . . to denote arbitrary formulas.

A statement like A ↔ B, that says that A and B are equivalent (“the same”)
can be written as (A→B) & (B→A).

A different view on mathematics: metamathematics

From the moment on that reasoning necessary for doing mathematics was cap-
tured by a few logical rules, the collection of mathematical theorems derivable
from some axioms became a well-defined total.

Axioms

Logic

Mathematics

Now the boundary around the mathematical theorems provable from the axioms
is quite precise. For this reason it is surrounded by a rectangle with sharp edges.
would be difficult to express. Having + and · as primitive operations the defining equations of
addition and multiplication need to be taken as axioms. In fact, the functional notation is so
convenient, that we assume that for many more functions f we have symbols in the language
of PA. This is justified since we can eliminate these symbols via the predicate Af as explained
above for addition.

36
 Now that the collection of mathematical theorems in an axiomatic system
is precisely defined, one may ask questions about this totality. For example the
following.

• Decidability.
Is it mechanically decidable (by machine, computer) whether a statement
is provable and hence a theorem?

• Completeness.
Is the collection of provable theorems the same as the true theorems? Or
avoiding the notion of ‘truth’:
Is for any statement A either provable or refutable16 ?

Leibniz had the optimistic hope to construct a machine that not only would
decide all mathematical statements, but even all philosophical ones. The first
question he wanted to ask to such a machine was: “Does God exist?” For
someone in the early 1700s he had a striking belief in technology. The mathe-
matician Hilbert had as belief that the axiomatic method was so powerful that
any statement could be settled, either in the positive or in the negative. Both
optimistic views turned out to be falls. Gödel’s incompleteness theorem (1931)
states that if an axiomatic theory is at least as strong as Peano arithmetic, then
if that theory is consistent, there are true statements that cannot be proved.
Turing’s undecidability theorem (1936) states that even if one does not assume
any axioms, it is not mechanically decidable whether a statement is provable
or not. For this he introduced the universal computer already encountered in
section 3.

Gödel’s theorem
The famous theorem of Gödel[1931] will be presented now. It states that if
Peano Arithmetic is consistent, then not all true statements can be formally
proved. This shows that the axiomatic method has limitations. But it is still
very good and in fact the best we have.
The method applies also to other axiomatic systems of arithmetic, provided
that they are at least as powerful as that of Peano. The outline of the reasoning
will be given first and then we enter some of the details. For a more complete
account of the proof see Nagel and Newman [2001] or Gödel’s original paper
cited below.

1. Arithmetical statements speak about numbers.

2. (Pythagoras) Everything is a number (after coding).

3. Arithmetical statements speak about everything you want (via coding).

4. Arithmetical statements speak about (other) arithmetical statements.

5. Some arithmetical statements speak about themselves (!).

16
A statement A is refutable if its negation ¬A is provable.

37
 6. L: This statement is false.

7. G: This statement is unprovable from the Peano axioms.

8. If PA is consistent (free from contradictions), then G is not provable and

hence true!

Conclusion. Interesting axiomatic systems (at least as strong as PA and

free from contradictions) are incomplete: there are statements G that are true
but not provable.

Now we discuss the details.

1. Arithmetical statements speak about numbers.

This is clear. A notion like “prime” can be expressed formally as

P (n) := (n > 1) & ∀d[d|n ⇒ (d = 1 ∨ d = n)].

Here we define the notion “d is a divisor of n”, notation d|n, by

d|n := ∃q d · q = n.

All formulas in PA speak about numbers.

2. Everything is a number (after coding).

At elementary school you may have invented a trick to code as numbers mes-
sages to your friends (and how to decode them). Currently and in past centuries
cryptography has become a hot topic, having military implications, see Singh
[2000]. A lot of mathematics is involved. But without special needs like secrecy,
coding is not very difficult. Pythagoras would have like the fact that a whole
symphony may be coded by one (large) number, as happens for example on a
music CD17 .

3. Arithmetical statements speak about everything you want (via coding).

One can define now formulas of PA such that e.g. F (x) states that the music
coded by x contains a flute passage, or if you prefer a passage of your favorate
singer.

These things were novel at the time that Gödel wrote his famous paper. Today
in the digital era, these facts are well-known. Now reflection becomes possible.

4. Arithmetical statements speak about (other) arithmetical statements.

Since an arithmetical statements is a member of the formal language form, it
may be coded as a number #A. Nothing special. Let A be the corresponding
numeral in PA. That is, if #(A) = 3, then A = S(S(S(0))).
17
If compactness is an issue, then things become harder. Also reliability is an issue: can
we reconstruct the music if we loose a bit of the digital information? MP3 is a standard for
coding music having both properties of compactness and reliability.

38
 For example one may construct a formula B(x) such that
Prov( A ) states that A is provable in PA.

5. Some arithmetical statements speak about themselves.

This is the essential step in the construction of self-reflection. One can construct
a function sx such that inside PA

sx ( A , n) = A[x := n] .

Here A[x := t] denotes the result of substituting the term t for the variable x
in A. Now define dx (n) = sx (n, n). Then inside PA one has

dx ( A ) = sx ( A , #(A)) = A[x:=#A] = A[x:= A ] .

Suppose we want to construct a PA formula “Self” that states that it, i.e.
“Self”, is provable. Then we can take

A(x) := Prov(dx (x))

Self := A[x := A ].

Indeed,

Self ↔ A[x := A ], by definition of Self,

↔ Prov(dx ( A )), by definition of A(x),
↔ Prov( A[x := A ] ), by the property of d,
↔ Prov( Self ), by definition of Self.

6. This statement is false.

Let us call this stament L. It states that L does not hold. Thus L is equivalent
with a statement like:
“I am lying now.”
If this statement is true, then I am lying indeed, but then (by definition of lying)
it should not be true. On the other hand, if it is false, then (by definition of lying
again) it is true. This is the famous liar paradox, already known to the Greek 18 .

18
Together with the construction in 5 it shows that there is no “truth predicate” T having
as property that
T( A ) ↔ A
holds in PA. Indeed, if such a T would exist, then one arrives at a contradiction by defining

L ↔ ¬T ( L ).

This result is due to A. Tarski. It show that decoding is not possible by one truth predicate T .
On the other hand, decoding in arithmetic is possible. But for this one needs to classify the
set of formulas in to classes Πn (basically indicating how many quantifier changes ∀∃ occur).
For these one has partial truth predicates Tn such that for all A∈Πn one has in PA

Tn ( A ) ↔ A.

The upshot is that reflection in the sence of section 1 is possible within PA, not by a single
mechanism but by a sequence of mechanisms.

39
 7. By 5. we can construct a formula G such that

G ↔ ¬Prov G .

8. Suppose that G is provable in PA. Then

Prov( G )

is provable in PA. But by the meaning of G also

¬Prov( G )

is provable in PA. This makes PA inconsistent. Therefore if PA is consistent,

then G cannot be provable. But then G is true, as it stated it own unprovability!

Conlcusion. If PA is consistent then it cannot prove the true statement G.

If PA is not consistent, then it can prove everything (including) G, but then
PA is not an interesting theory.
The reasoning can be done for every axiomatic theory T at least as strong
as PA19 .

Theorem. Every consistent theory T as least as strong as PA contains a true

but unprovable statement GT .

This result, obtained via reflection, shows the limitation of the axiomatic
method. The situation is not terribly bad, however. The axiomatic method is
still quite powerful. Some people conclude that the human mind has an essen-
tially non-mechanical basis. The reasoning is: “We humans can conclude that
GT is true, but the (mechanical) axiomatic system cannot.” See Penrose [1989],
[1994]. We disagree, perhaps not with the conclusion,but with this argumenta-
tion. The validity of GT depends on the assumption of the consistency of T.
And it is not clear that we can obtain the insight that this is the case. A lucid
description of the many theories of mind can be found in Blackmore [2004].
Gödel’s theorem states that arithmetic truth and arithmetic provability are
not the same. Sharpening Gödel’s reasoning Rosser showed that if PA is con-
sistent the Gödel sentence can neither be proved nor refuted in this theory.

Undecidablity
The following theorem of Church makes another metamathematical statement
about arithmetic.

Theorem. (Church) Let T be a consistent extension of Peano Arithmetic.

Then there is no computable decision method to determine whether a state-
ment A in the language of arithmetic belongs to T .

19
For this one assumes that “axiomatic” means that the axioms can be recognized as such.
But that is a reasonable assumption.

40
 Similar but for a simpler theory is the following theorem of Turing.

Theorem. (Turing) There is no computable decision method to determine

whether a statement A of Frege’s logic is provable in this theory.

This result shows that the ideal of Leibniz to construct a machine that could
answer all precisely stated questions cannot be fulfilled.

For details of these theorems see Davis [1965].

For some axiomatic theories there does exist, however, a computable deci-
sion method.

Theorem (Tarski) The theory of elementary geometry has a computable deci-

sion method.

The technical proof may be found in Tarski [1951].

Exercises
4.1. Show that ∀x∀y (x + y) = (y + x) by induction ‘on y’.
4.2. Define

x0 = 1;
xn+1 = xn .x.

Show by induction on n that for all x

(1 + x)n ≥ 1 + nx.

4.3. Define an arithmetic predicate Q such that for terms t one has

Q[x := t ] holds if and only if t starts with a 0.

References

Aristotle [350 B.C.]. Organon, See also <classics.mit.edu/Aristotle>.

Blackmore, S. [2004]. Consciousness, an Introduction, Oxford University Press,

Oxford.

van Dalen, Dirk [1994]. Logic and structure, Universitext, third edition,
Springer-Verlag, Berlin.

Davis, M. (ed.) [1965]. The undecidable. Basic papers on undecidable propo-

sitions, unsolvable problems and computable functions, Raven Press,
Hewlett, N.Y.

41
Euclid [300 BC]. Elements, Green Lion Press, Santa Fe, NM. All thirteen
books complete in one volume, The Thomas L. Heath translation, Edited
by Dana Densmore, 2002.

Gödel, K. [1931]. Über formal unentscheidbare Sätze der Principia Mathemat-

ica und verwandter Systeme, Monatshefte für Mathematik und Physik 38,
pp. 173–198. Translated and commented in Gödel [1986]. Another English
version based on course notes by Kleene and Rosser is in Gödel [1965].

Gödel, Kurt [1965]. On undecidable propositions of formal mathematical sys-

tems, in: Martin Davis (ed.), The Undecidable: Basic Papers on Undecid-
able Propositions, Unsolvable Problems and Computable Functions, Raven
Press, New York, pp. 41–74. From mimeographed notes on lectures given
by Gödel in 1934.

Gödel, Kurt [1986]. Collected works. Vol. I, The Clarendon Press Oxford Uni-
versity Press, New York. Publications 1929–1936, Edited and with a pref-
ace by Solomon Feferman.

Hilbert, David [1899]. Grundlagen der geometrie, Festschrift zur Feier der
Enthullung des Gauss-Weber Denkmals in Göttingen, Teubner, Leipzig.
Later editions published in 1903, 1909, 1911, 1922 and 1930. Translation
by Leo Unger of the 10th German Edition: Foundations of Geometry;
Open Court, LaSalle, 1971.

Nagel, Ernest and James R. Newman [2001]. Gödel’s proof, revised edition,
New York University Press, New York. Edited and with a new foreword
by Douglas R. Hofstadter.

Penrose, Roger [1989]. The emperor’s new mind, The Clarendon Press Oxford
University Press, New York. Concerning computers, minds, and the laws
of physics, With a foreword by Martin Gardner.

Penrose, Roger [1994]. Shadows of the mind, Oxford University Press, Oxford.
A search for the missing science of consciousness.

Singh, S. [2000]. The Code Book, Anchor Books.

Tarski, A. [1951]. Decision Method for Elementary Algebra and Geometry,

University of California Press, Berkeley.

42
5. Reflection and Art

Text by Prof.dr. Anneke Smelik.

To be given during the course.

43
6. Reflection and Computers

What computers can’t do

by Mike Yates20
http://plus.maths.org/issue5/turing

Alan Turing is described by Professor P.N. Furbank, overall editor of Turing’s

Collected Works[1], as ”one of the leading figures of twentieth-century science”.

Figure 6: Alan Turing.

Sixty years ago his most famous paper was published, introducing the idea of
a Universal Computing Machine ten years before the first stored programme
digital computer actually ran.
This was only one of a string of varied achievements. It is known now that
his work on deciphering the German Enigma code at Bletchley Park during the
Second World War made a significant contribution to winning that war, though
this remained unknown to his closest friends until after his tragic death from
taking potassium cyanide in 1954.

Figure 7: The Enigma machine21 .

20
Emeritus Professor of the University of Manchester, and an Honorary Professor of the
University of Wales at Bangor.
21
Image source : AGN, University of Hamburg, Copyright 1995, Morton Swimmer.

44
 Turing’s wartime work played a significant role in marking out the impor-
tance of mechanical computing facilities. Although much of the hack work was
done mechanically, an enormous team of human computers was also involved.

Figure 8: Close-up of the coding rotors.

Another feature of his wartime work was its use of probability theory. Some
of Turing’s work in this area was also highly innovative. It was recognized after
the war through the published work of his then assistant (later Professor) Jack
Good, without reference to its wartime uses.
Turing’s interest in computing continued after the war, when he worked at
NPL (National Physical Laboratory) on the development of a stored-programme
computer (the ACE or Automatic Computing Engine). In 1948 he moved to
Manchester, where the first stored programme digital computer actually ran
that year.
Although his connection with that real computer was at best tenuous, he
made significant contributions to computing theory, in particular artificial in-
telligence (the Turing test), computer architecture (the ACE) and software
engineering. It is some measure of his contribution that the prestigious Turing
Prize in computing science is named after him.

Figure 9: In the Turing test for machine intelligence, an observer has to distin-
guish between the machine and a human by asking a series of questions through
a computer link.
In the Turing test for machine intelligence, an observer has to distinguish

45
between the machine and a human by asking a series of questions through a
computer link.

The halting problem

As an example of his thought let’s look at a proof that there is no way of
telling in general once a computer has embarked on a calculation whether that
calculation will terminate in an answer. This problem is known as the ”Halting
Problem for Turing machines” and was first proved in the 1937 paper[2] in
which he introduced his machines.
To lead up to that proof, it is necessary to say a few things about counting
and lists or sequences. We say that the elements of a set can be counted if they
can be listed in a single sequence.
The set of natural numbers can be listed 0, 1, 2, 3,... and so on ad infinitum
- no problem. To list all the integers, positive and negative in a single sequence,
you can write 0, 1, -1, 2, -2, 3, -3,... and so on, again no problem.
The fractions take a bit more work. It is usual to do this in 2D, using a
table or matrix. Let’s just look at the positive ones - it extends to include the
negative ones as with the integers.

Figure 10: Table of fractions.

The fractions can be counted by tabulating them and then counting them along
the diagonals, shown in blue.
There are a lot of repetitions - all the diagonal elements are equal for a
start - so this algorithm is a little wasteful. But it does the job. Carry it on for
ever and every fraction will be there somewhere in the 2D matrix. To write the
matrix out in a single sequence, work up and down the SW to NE diagonals to
obtain:

1, 1/2, 2, 1/3, 2/2, 3, 1/4, 2/3, 3/2, 4,...

Next, we come to a very famous theorem, Cantor’s Theorem, which says that
the real numbers are not countable in this way. The set of real numbers include
numbers like π (=3.14159...) which cannot be written as one whole number
over another.

46
Proof of Cantor’s Theorem
Let’s just show that we cannot count all the binary sequences, in other words,
infinite sequences of 0s and 1s.
Suppose we could. We can label each binary sequence B1, B2, B3,... ad
infinitum. We will now obtain a contradiction. Let’s list the elements of each
sequence in a table or matrix as before.

Figure 11: Table of binary sequences. A possible list of binary sequences, the
sequence D is constructed by inverting the items on the diagonal, shown in blue.
Now define a binary sequence, D, by choosing a 0 in the first column if B1
has a 1 in that column and 1 if B1 has a 0 in that column. We then choose
a 0 in the second column if B2 has a 1 in that column and 1 if it has a 0 and
so on. The resulting binary sequence, D, cannot be in the list because if it
were it would have to match one of the B sequences, say Bn for some n. But
we have just deliberately made sure that the nth column of D differs from Bn.
Contradiction.
No matter how we list the binary sequences we can always find a new se-
quence, D, which is not in the list.
This procedure is called diagonalizing. As you can see, we have given a
simple rule for it, so that given a rule for counting out a list of binary numbers
then we’d have a rule for computing this diagonal binary number which isn’t
in the list.

Turing’s argument
Finally, let’s sketch how Turing’s argument (related to an even more famous bit
of reasoning by Kurt Gödel in 1931) takes this argument a big stage further.
The proof sketched here is not Turing’s original one, but related. Much of
Turing’s classic paper is taken up with describing his concept of a computing
machine and why it is as general as can be. Anything that can be computed
according to a finite list of rules, can be computed by one of his machines.
Briefly, a Turing machine can be thought of as a black box, which performs
a calculation of some kind on an input number. If the calculation reaches a
conclusion, or halts then an output number is returned. Otherwise, the machine
theoretically just carries on forever. There are an infinite number of Turing

47
machines, as there are an infinite number of calculations that can be done with
a finite list of rules.
One of the consequences of Turing’s theory is that there is a Universal
Turing machine, in other words one which can simulate all possible Turing
machines. This means that we can think of the Turing machines as countable
and listed T1, T2,... by a Universal Machine through a sort of alphabetical
listing. Turing used this to describe his own version of Gödel’s Theorem: that
there is no mechanical procedure for telling whether a Turing machine will halt
on a given input: the Halting Problem. The unsolvability of the halting problem
Let’s represent the result of using the nth Turing machine, Tn on the input
i as Tn(i). Suppose that there was a rule or procedure for deciding whether or
not Tn(i) halts for all values of n and i.

Figure 12: A halting rule could be used to make a table of the output Tn(i),
using a question mark to represent calculations which never halt. This table is
only illustrative, its contents have not been chosen with any particular ordering
of Turing machines in mind.
But then by a similar diagonalizing procedure to the one above, we can
define a new Turing machine, say D, which will halt for all inputs and return
the following output for input i:
0 if Ti(i) does not halt. Ti(i)+1 if Ti(i) does halt.
But this machine D must be one of those machines, in other words it must
be Td for some d. However, we just defined it to give a different answer from
Td with input d. Contradiction.
The extra sophistication here over the original diagonalizing argument lies in
(1) all the listing done is itself computable and (2) any machine Tn may or may
not halt in carrying out its computations. None of this enters into Cantor’s
original diagonal argument. This sort of computable diagonalizing was first
used in the pioneering work done by Gödel, Turing and others in the decade
before the Second World War, and has remained an important technique. The
really hard work lies in formulating the various definitions of computability, but
that is another story! What is life?

48
 Figure 13: Turing’s meticulously hand-drawn sunflower.
In the closing years before his death, Turing was working on something
entirely different, something which had been close to his heart since his school
days - the origin of biological form - Morphogenesis.

Figure 14: A close-up section.

How could simple cells know how to grow into relatively enormous structured
forms? The crucial idea that genetic information could be stored at molecular
level had been deduced in Schrdinger’s 1943 lecture What is Life?, and Crick
and Watson were currently busy in the uncovering of that secret, through the
structure of DNA. Given the production of molecules by the genes, Turing was
looking for an explanation of how a chemical soup could possibly give rise to a
biological pattern.
The first main goal of his theory was an attempt on the classic problem of
Phyllotaxis, the arrangement of leaves on a plant. One of the features of this
subject which had been known since Kepler’s time was the natural occurrence
of the Fibonacci series 1, 2, 3, 5, 8, 13, 21,... So it was already established that
mathematics had a role to play. (For more about the Fibonacci series see ” The
life and numbers of Fibonacci” in Issue No 3.)
Turing also proposed that the pattern of markings on animals followed math-
ematical rules due to chemical signals. This idea had mixed fortunes, though
recently biologists’ interest has been re-vitalised. Using his theory, researchers
in Japan have observed Turing’s predicted changes in the patterns on zebra-
striped fish.

49
Further reading
Glance at the web page

http://www.turing.org.uk/sources/biblio.html
(The Alan Turing Bibliography, assembled by Andrew Hodges) for further de-
tails of the Collected Works.

The definitive work on his life (a compelling read) is:

• Andrew Hodges, Alan Turing: The Enigma, hardback version - Burnett

books, 1983, paperback version - Vintage Books, 1992.

A new angle on Turing can be found in:

• Andrew Hodges, Turing, in the Series The Great Philosophers, Phoenix

1997.
A guiding force from his school-days was:

• D’Arcy Wentworth Thompson, On growth and Form, Cambridge Univer-

sity Press. 1917 (new edition 1942).

References
[1] Collected Works of A. M. Turing edited by P.N. Furbank, North-Holland,
Elsevier Science.

[2] Turing, A. On Computable Numbers, with an application to the Entschei-

dungsproblem, Proceedings London Mathematical Society (series 2) vol 42,
1936-7, pp. 230-265.

-||-

A very readable history of the creation of the computer is the following.

Davis, M. Engines of Logic, Mathematicians and the Origin of the Computer,

Norton, 2000.

The classical papers on the theory of computability can be found here.

Davis, M. The undecidable. Basic papers on undecidable propositions, unsolv-

able problems and computable functions, Raven Press, 1965.

50
7. Reflection and the human mind

Happiness and suffering are both the result of two factors combined: the situ-
ation in which one is placed and our consciousness of it. Happiness is not only
of personal importance, it is also a necessary factor for ensuring peace in soci-
ety. Therefore it is important to know the two possible ways for the pursuit of
happiness: applied science, which focuses on how situations can be controlled,
and spirituality, which focuses on developing the various types of consciousness
one can have.
Usually we say that we have a body. Our body may sit in a certain position.
Or it is moving around. But is it really ours? When we are ill we prefer not to
have this body. When we are healthy it needs food. We are not in full control
of our body.
The same may be applied to our mind. Sometimes when we have decided
to stay in our room in order to study, the mind gets excited and wants us to
go out. If we have the discipline to follow our plan to study our excitement
sucks and we can study less. Alternatively, if we give in and go out, then our
guilt about not studying may spoil our evening. We see that we are even less
in control over our mind.

Says Saint Augustine:

If my mind orders my body to do something,
then my body obeys so well,
that one can hardly distinguish between the order and its execution.
If, however, my mind orders my mind to do something,
then it does not listen, even if it is the same mind.
Why this monstrous phenomenon and for what purpose?
Schopenhauer says something related:
We are able to want to do something,
but we cannot [force ourselves to] want to want something.
This section on the mind will describe a way to use reflection in order to
gain a higher degree of mental freedom. The method is handed down from
classical Buddhism. I do not want to claim that it is the only way to reach this
freedom. But the the freedom obtained essentially depends on understanding
the functioning of the mind.

Investigating the mind

If we find ourselves in a situation that is not agreeable to us, there are essentially
two different ways to attempt to overcome the resulting suffering. On the one
hand we can try to change the situation itself. On the other hand we can
try to change our consciousness of the situation in such a way that it is no
longer experienced as unpleasant. Depending on the circumstances and our
possibilities, these attempts will be more or less successful.

51
 It is clear that on the one hand science, technology and democracy have
contributed considerably to the way our world can be controlled. On the other
hand, we have Buddhism, in particular the ‘theravada’ school (literary the
‘teachings of the elder’) that is focussed on the second possibility: eliminating
suffering by changing our consciousness. Having completed successfully the
course described above implies that one is accomplished in obtaining a different
view on the things that happen to us. This state of mind is called equanimity
and should not be confused with indifference. It may take many years to reach
a steady form of equanimity, but the aim is worthwhile. It does not imply that
one becomes passive, quite on the contrary.
In order to decrease conditioning of the mind and hence the resulting suf-
fering one needs (intuitive) insight. This is not insight through intellectual
understanding. The difference is this. When one learns to ride a bike, then one
knows that making a turn one needs to bend over. Using mechanics as laid out
by the laws of Newton one may calculate the possible angles that are needed to
make a smooth bent. But being able to do this, this does not necessarily mean
that can ride a bike. The way a child learns to ride is an intuitive one. After a
couple of trials he or she intuitively knows how to make a smooth bent.
The insight needed for understanding our mind requires another mental
faculty. This is concentration. It is interesting, however, that if one wants to
be concentrated then one cannot always be in that state. Even if someone
would pay you some good money to be concentrated for one hour, you may not
succeed.
Nevertheless, concentration can be developed. For this one needs to have
some discipline and this finally one can make oneself to do this. For example,
you all came to class and this requires some discipline that you have been willing
to put energy into.
Using discipline, concentration one can develop insight, but not by forcing
it. Like in the sports and in music playing one can reach a certain level by
cultivating our possibilities. Insight that is obtained will have its positive effect
on discipline and concentration, by seeing what are factors that disturb us an
finding ways to avoid them. This growing process with its feedback is all part
of mental development.

Mindfulness
In the presence of discipline and concentration one is able to develop insight.
This is done by applying mindfulness. Usually our senses get a lot of input from
the world. Mindfulness is the observation of the mental phenomena. When
someone calls you a bad name, like ‘monster’, then you may feel offended and
react accordingly. Using mindfulness one first observes the hearing of the word,
and then observes the reaction of feeling offended, and then observes the reac-
tion of angriness or perhaps sadness. Mindfulness creates a distance between
oneself and the phenomena. One is not sucked away in the stream of conscious-
ness as one usually is.
In order to be able to do this one needs exercising. Sitting meditation is an
exercise to increase mindfulness by focusing on breathing (raising and falling

52
of the abdomen). Walking meditation is an exercise in mindfulness on our left
and right footsteps. The exercises are done one after the other many times, for
say half an hour each.
It is this mindfulness that gives intuitive insights. The cogito of Descartes
can be seen as an act of mindfulness giving rise to insight. Descartes was
wrestling with the question what is the nature of the mind. ‘Do I really exist?
There is matter. But is there mind? What can one be sure of?’ Deliberating in
similar ways Descartes suddenly observed with mindfulness the act of wrestling
thoughts, cognitions. Then he shouted: ‘Cogito ergo sum!’ (‘I deliberate hence
I am!’) This ‘Cogito’ should not be seen as the intellectual thinking, but as the
mindful observation of the thinking, which act belongs to intuitive insight. By
seeing himself from a different level he experienced existence more than he had
done hitherto.

The three characteristics

Now we go into the details of the kind of insights one may obtain through
mindfulness based on discipline and concentration. If we hear someone say
to us ‘monster’, then there is hearing. Usually this hearing is blended with
feeling. Mindful observation shows on an intuitive level, that the sound and the
emotional contents are different. Rationally this is very easy to understand.
To separate the two components on an intuitive level is more difficult. Insight
works as a dissecting knife and separates mental phenomena in more primitive
forms: the input (through the senses), the feeling (giving value judgments: to
obtain, to avoid, to keep), cognizing (distinguishing this from that) and output
(reacting through the muscles and inner states).
Our usual view of ourselves as a constant agent that acts in the world
will eventually get transformed. We do not have a steady ‘soul’ but consist
of a bundle of phenomena. When we realize this on the intuitive level, we
see the three properties of life: everything is subject to a constant change, is
unsatisfactory and is beyond control. In the Buddhist tradition22 from which
insight meditation comes these are called the three fundamental characteristics:
non-permanence (anicca), suffering (dukkha) and non-self (anatta).
Experiencing the three characteristics is a powerful happening. Although
the view of ourselves as a bundle of phenomena is theoretically both under-
standable and even plausible, as actual experience it is something we would
like to avoid at all costs. Continuing the investigation of the mind, we see that
we usually are covering-up the three characteristics by hiding them with our
feelings. As soon as the cover-up wears off, we have to find a new way to hide
the characteristics. The way we do this depends on our personality. We can
be a femme fatal or a wall flower, a Tarzan or an underdog, just to mention a
few possibilities, or anything in between or combined. Each personality has a
22
This is Theravada Buddhism as practised in Myanmar, Thailand and Sri Lanka. Its
vipassana (insight) meditation is at present widespread in the West. In 2003 there was a
program on Dutch TV (Netwerk how vipassana was used in prisons. Mindfulness based
cognitive therapy, see Segal et al. [2002], is a psychotherapeutic application of the method of
mental development through insight.

53
private way to create feelings to glue the bundle of phenomena together. This
gluing together is a symptomatic treatment and makes us slaves of our habits.
See Barendregt [1988] for a personal account of the development of mindfulness
and the encounter with the three characteristics.

Purification
Mindfulness, that so far has been used as a tool to obtain insight, now can be
used as a way to purify the mind. A continuous and concentrated application
of mindfulness gives a stable consciousness, one of which one does not become
dependent. Here ‘concentrated’ means that one performs the act of mindfulness
as often as possible; ‘continuous’ means that one does this for prolonged periods.
When one is able to do this, temporary happiness is obtained. But purifi-
cation still has not taken place. Indeed, one has to make mindfulness. This
requires energy and concentration. The application of non-interfering mindful-
ness has indeed a purifying effect. But the effect is not lasting, as mindfulness
comes in waves that depend on how we are pushed by our personality. By
discipline and concentration the continued concentrated mindfulness should be
maintained as long as is possible and in a comfortable way.
This needs to be trained. When this can be done together with calmness,
equanimity and bliss one has to surrender. Then the mindfulness may be-
come permanent and automatic. The main bottleneck is to become ready for
this. Indeed, letting go is usually coupled with the phenomenon of fear, since
humans have the urge of being always in control. Another difficulty to over-
come is caused by the tendency to be sidetracked by euphoric experiences. See
Barendregt [1996] for a personal account of this process.

Mindfulness and reflection

The essence of mindfulness is attention with detachment. This detachment may
be compared to going to a ‘meta-level’. Consider the following notions from
language semantics. There are objects in reality and names in language.

Reality Language
1 (:- ))|-< Maria

2 Maria ‘Maria’
Figure 15: Naming

In line 1 of Fig. 15 under ‘Reality’ we see (the picture of) a (lying) girl; under
‘Language’ we read her name. In line 2 we see that the name of the girl also
occurs in reality; and that the name in quotes is the name of the name. One
can state: “Maria is a nice girl” and “The name ‘Maria’ consists of five letters.”

This phenomenon may be compared to the act of mindfulness. The mind is

always capturing an object (for example our breathing). This is shown in line
1 of Fig.16, where under ‘Object’ a process is described that is observed by con-

54
 Object Mind
1 breathing mental event

2 mental event mindfulness

Figure 16: Mindfulness

sciousness, in this case the observed object is the physical raising and falling
of the abdomen. Under ‘Mind’ the inner awareness of the object is denoted.
This awareness of the object in our consciousness can be ‘recollected’ (this is
the literal meaning of sati, the ancient Pali23 word for ‘mindfulness’). In that
sense the mental event becomes an object for the mind and its awareness is
mindfulness, see line 2 of Fig. 16.
Using mindfulness one works, in computer science terms, with the code of
consciousness rather than with its executable. Therefore one is detached without
loosing any information. In this way one can react in an equanimous way to
phenomena that are ‘as if’ desire or ‘as if’ fear. This form of reflection gives
pure consciousness.
The path of purification using mindfulness has been traversed traditionally
in the monastic tradition. A monastery is the ‘laboratory for the mind’. This is
how a vipassana teacher at a meditation retreat may speak to the meditators.

“Thou dwellers of the great monastery: work with confidence, understanding,

effort, concentration and above all mindfulness. At first restrain your senses
and stay with their input as much as possible. Make a mental note if your
consciousness is pulled elsewhere. This eventually will set you free and your
sensory restraint has served its purpose. Be aware of two pitfalls. Too much
concentration may give apparent freedom; but you will fall back. Secondly, it is
not you who can finish the work. Start with your desire to be unconditioned. At
some point you will see that it reaches nowhere. Then let discipline take over
and surrender with attention. Do not expect anything and the work will be over
soon: bliss of Nibbana24 becomes permanently accessible. In this life you may
use it for the benefit of all living beings.”

In the Buddhist tradition it is recognized that the path of purification, can be

walked by monks but also by lay people.

Psychotherapy
Partial mindfulness may happen in daily life can reveal a glimpse of the three
characteristics. The resulting view of the unsatisfactoriness of our consciousness
as loosely bound bundle may create some mental unbalance or hackneyed cover-
up as we find in fobias and depression. Mindfulness based stress reduction
(MBSR) and mindfulness based cognitive therapy (MBCT) have been developed
to treat these. See Kabat-Zinn [1990] and Segal et al. [2002].
23
The language in which the ancient theory of mind is written.
24
Pali for Nirvana, the state of pure consciousness.

55
References
Barendregt, H.P.
[1988] Buddhist Phenomenology, in: Proceedings of the Conference on
Topics and Perspectives of Contemporary Logic and Philosophy of Science,
Cesena, Italy, January 7-10, 1987, (Ed. M. dalla Chiara), Clueb, Bologna,
1988, 37-55. Available at <www.cs.kun.nl/henk/BP/bp1.html>.

[1996] Mysticism and Beyond, Buddhist Phenomenology, Part II,

The Eastern Buddhist, New Series, vol. XXIX (1996), pp. 262 287.
Available at <www.cs.kun.nl/henk/BP/bp2.html>.

Goldstein, J. & J. Kornfield

[2001] Seeking the heart of Wisdom: The Path of Insight Meditation,
Shambala Classics.

Kabat-Zinn, J.
[1990] Full Catastrophy Living: The program of the Stress Reduction Clinic
at the University of Massachusetts Medical Center, New York, Delta, 1990.

Segal, Z.V., J.D. Teasdale and J.M.G. Williams,

[2002] Mindfulness-Based Cognitive Therapy for Depression: A new
approach to preventing relapse. New York, Guilford Press

56