Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Edi t or and Founder :

Wi l l i amM. Honi g
Publ i shed f i ve t i mes per annum
I SSN0155-7785
Vol ume6 Number 2
J une 1983
El sevi er Sequoi a -
Lausanne
Specul at i ons i n
Sci ence andTechnol ogy, Vol . 6, No. 2( 1983) - p
. 103- 112

( 47L) 103
THEGENESI S MODEL - PARTI I :
FREQUENCYDI STRI BUTI ONS
OFELEMENTSI N
SELFORGANI ZEDSYSTEMS
PETERWI NI WARTER
Le Bor dal i er I nst i t ut e, F- 41270 Dr oue,
Fr ance,
Recei ved: 7Sept ember 1982
Abst r act
Aquant i t at i ve measur ement of t he compl exi t y Cof a sel f or gani zed syst emi mpl i es t he
det er mi nat i on of t he f r equenci es of t he el ement t ypes const i t ut i ng t he syst em. The anal ysi s
of empi r i cal dat a concer ni ng a l ar ge var i et y of sel f or gani zed syst ems - such as cosmi c
syst ems, bi ol ogi cal syst ems, economi c syst ems and l i ngui st i c syst ems - suggest s a gener al
l aw. The secondl awof genesi s:
The
r ank- or der ed
f r equency
di st r i but i on of el ement - t ypes of
any sel f or gani zed syst emat any l evel of hi er ar chi cal or gani zat i on can be appr oxi mat edby
a mat hemat i cal di st r i but i on of t he f or m
pn
=A( n
+
m) - B-
1. I NTRODUCTI ON
I n par t I of t he genesi s
model ")
we have i nt r oduced a quant i t at i ve
measur e f or t he compl exi t y of a sel f or gani zed syst emof mat t er C=I R. The
post ul at ed f i r st l awof genesi s AC>
0
i mpl i es, t hat f or a syst emwi t h const ant
ener gy r edundancy Rt he aver age sel ect i ve i nf or mat i on per ener gy st at e
i ncr eases ( AI >0) and t he syst emi s at equi l i br i umi f a maxi mumval ue I , nax
cor r espondi ng t o t he maxi mumval ue of t he ent r opy
Smaa
i s r eached.
Let us nowconsi der t he l ogi cal count er par t , t hat i s consi der asyst emwi t h
a const ant val ue of I . Appl yi ng t he f i r st l awof genesi s AC>0 i mpl i es AR>0.
Such t he ener gy r edundancy of a syst emwi t h const ant i nf or mat i on shoul d
i ncr ease and t he syst emi s at equi l i br i um, i f a maxi mumval ue R, nax i s r eached.
I n anal ogy t o Bol t zmann' s appr oach i t woul d be i nt er est i ng t o cal cul at e
t he t heor et i cal f r equency di st r i but i on of el ement s pi cor r espondi ng t o
Rmax
and a gi ven val ue of I , usi ng t he met hod of Lagr ange mul t i pl i er s. Thi s woul d
i mpl y t hat t he f unct i onal r el at i onshi p of t he quant i t y t o be maxi mi zedRand
t he di f f er ent r el at i ve f r equenci es pi be known expl i ci t l y
.
I n our def i ni t i on of
t he ener gy r edundancy
M

M
R=
[( Z
ni ch, - E. )
/ E
ni ei q
]
X
100
[%]
i =>

i =>
wher e ni t he absol ut e f r equency of t he el ement - t ype wi t h t he ener gy at r est
pr r est - mass
ei q , andEq
t he
ener gy
at
r est or r est - mass of t he t ot al syst em, t he
dependence of En on t he di st r i but i on ni or pi =n; / Ni s not known expl i ci t l y.
Leavi ng asi de t hi s t heor et i cal hur dl e,
we
have chosen a pat h demandi ng
"l ess ef f or t " by l i mi t i ng our sel ves t o t he anal ysi s of empi r i cal dat a.
El sevi er sequoi a S. A. , Lausanne. Pr i nt edi n
t he Net her l ands.
0155- 7785/ 83/ $3. 00
104 ( 48L)

Pet er Wi ni war t er
2.

ZI PF' S ANALYSI SOFHUMANLANGUAGE
"Human Behavi our
and
t he
Pr i nci pl e of Least Ef f or t " i s t he t i t l e of a
st udy of Zi pf l 2 i publ i shed
i n
1949. We i gnor e
whet her Shannon' s andWeaver ' s
"Mat hemat i cal Theor y of
Communi cat i on"(
3
)
publ i shed i n book f or mi n t he
same year have i nf l uenced
t he wor ks of Zi pf , but we suppose t hat - as so of t en
i n t he hi st or y of sci ence -
t he di scover y of Zi pf ' s empi r i cal l awand
Shannon' s
quant i t at i ve def i ni t i on of
sel ect i ve i nf or mat i on have evol ved i ndependent l y
.
Asummar y of
Zi pf ' s t edi ous but r emar kabl e st udy r el evant t o
t he
i deas
pr oposedi n t hi s paper i s
gi ven byCher r y") ;
"Fi gur e 1 shows
cur ve A, t he r esul t of st at i st i cal anal ysi s
made upon
J ames J oyce' s Ul ysses
; t he vol ume cont ai ns about a quar t er of a mi l l i on
wor d
t okens wi t h a vocabul ar y of near l y 3 0, 000 wor d-t ypes. ( Token i s t he name of
ever y i ndi vi dual wor dt hat act ual l y appear s
i n pr i nt ed
t ext ;
t ype r ef er s
t o
t he
ent r i es of a vocabul ar y l i st or di ct i onar y of t he t ext . ) Thi s
cur ve
A
r esul t s f r om
pl ot t i ng t he f r equenci es of t he var i ous wor d-t ypes agai nst t hei r r ank-or der . ( I n
st at i st i cal st udi es, i f a number of el ement s ar e l i st ed i n decr easi ng or der of
t hei r
f r equenci es of occur r ence f , , f 2 . . . f n . . . t hen t hey
ar e
sai d
t o
be r ank-or der ed
i n f r equency. The suf f i xes 1, 2. . . n . . . may be r egar ded as uni t s on
a
l i near
scal e of r ank-or der . ) Sever al aspect s of t hi s cur ve
ar e
r emar kabl e.
Nat ur al l y,
t hi s cur ve must sl ope downwar d f r oml ef t t o r i ght , but we have
no
r i ght
what soever t o assume t hat any par t of i t woul d be at al l smoot h -
l et
al one
st r ai ght . I t mi ght wel l descend f r oml ef t t o
r i ght
i n
a ser i es of i r r egul ar j umps ;
agai n, r at her t han appr oach a st r ai ght l i ne, i t mi ght
t ake t he f or mof a dot t ed
cur ve as Cor D.
Fi gur e 1.

The
r ank-f r equency
di st r i but i on of wor ds
: ( A)
J ames J oyce' s "Ul ysses";
( B)
Amer i can newspaper
Engl i sh
; ( C)
and
( D)
hypot het i cal ( af t er Zi pf ) .
TheGenesi s Model - - Part 11

(49L) 105
Such a
l i near l awi s deri vedf romempi ri cal dat a; i f t he source of dat a be
changed markedl y, i t may
be f el t t hat t he change woul d be ref l ect edi n t he
f ormof t he l aw.
But Zi pf t akes some di f f erent dat a, correspondi ng t osampl es
of Ameri can newspapers,
and pl ot s t hemas i n curve B. Consi deri ng t he
di vergent nat ures of t hese sources of
l anguage, t he t wocurves AandBare
surpri si ngl y si mi l ar . Zi pf
rei nf orces hi s evi dence f or t he exi st ence of a def i ni t e
"l aw" by amassi ng si mi l ar dat a f romwi del y
di f f erent l anguages of t he worl d
(f or exampl e, see Fi gure 2bel ow), andf romt ext s coveri ng a t housand
years of
hi st ory. Not onl y words but ot her segment s
of t ext have been st udi edi n such
a st at i st i cal manner ; phonemes, syl l abl es,
morphemes - and even Chi nese
charact ers, andt he babbl i ng of babi es . "
11
RANK n 10 100 1000

10 . 000
Fi gure 2. The rank- f requency di st ri but i on of words: A- CNorwegi an; NGerman (af t er Zi pf )
.
3.

MANDELBROT' SEXPLI CATI ONOF ZI PF' S LAW
I n a det ai l edst udy Mandel brot l
s
) shows t hat Zi pf ' s empi ri cal l aw
pn =
AnB
pn =A(n +m)- B

r`

(2)
where

A=1/ E(n
+
m)- B
n
where AandB
are const ant s andB=
1
or a
l aw,
whi ch approxi mat es t he
empi ri cal dat a
even
bet t er
B=const . , andmi s a paramet er, can be deri vedas t he resul t of an opt i mi za-
t i on maki ng t hef ol l owi ne assumpt i ons:
106( 50L)

Pet er Wi ni wart er
( a)
( b)
The number
of l et t ers M, and
t he number of pot ent i al
words Mw are
const ant .
The average
sel ect i ve i nf ormat i on
per word of t he
consi dered syst emi s
gi ven i n
advanceandconst ant
.
1 =-E
pn
l og,
pn
=const.
n
( c)

The
quant i t y t o be
opt i mi zed/ mi ni mi zed i s
t he "average cost per word"
of t he
syst em:
Z f i n en
n
where cn i s a hypot het i cal
"cost " of t he
word-t ype occurri ng wi t h
t he
f requency
pn
measured i n
arbi t rary uni t s
of ent i ers .
Assumi ng t hat t he
"cost " of a word depends
on t he "cost s" of i t s con-
st i t ut i ng
l et t ers onl y, and
maki ng vari ous
assumpt i ons about t he "cost s"
assi sgned t o
l et t ers ( al l l et t ers of
equal cost ,
any cost assi gned t o t he vari ous
l et t ers of
t he al phabet -or
t he cost of anyl et t er i n a
word dependi ng upon t he
precedi ng
l et t er) Mandel brot
shows, t hat i n al l cases
t he cost of t he n-t h
group
of l et t ers
ordered accordi ng t o
i ncreasi ng "cost "
can be general l y expressed
i n t he
f orm:
cn - ca +
109m
x ( n +
m)
The met hod
of Langrange mul t i pl i ers yi el ds
f or t he mi ni mumof t he
"average cost per word" c
gi ven const ant Ml , Mw andI a
di st ri but i on of t he
f orm:
pn

=
A
eB
<n

( 4)
Repl aci ng cn i n ( 4) by expressi on
( 3) yi el ds a di st ri but i on
whi ch can be
approxi mat ed by ( 2) .
Such i n bi l ogari t hmi c coordi nat es
t he expressi on
-l og
pn
=
-l og A+Bl og( n +
m) descri bes most of Zi pf ' s
empi ri cal dat a i n a
sat i sf yi ng
way.
Mandel brot ' s approach
i s very cl ose t o t he general
probl emposed i n
sect i on 1, t he onl y di f f erence
bei ng t hat we want ed t o det ermi ne
t he f requency
di st ri but i on
p
n
yi el di ng a
maxi mumof t he
energy redundancy Rof a syst em
wi t hgi ven
const ant I , whi l e Mandel brot
det ermi nes t hef requency di st ri but i on
yi el di ng a mi ni mumof
t he average hypot het i cal
"cost " per word
f or
a
syst em
wi t h gi ven const ant
I .
Assumi ng
t hat Mandel brot ' s hypot het i cal
"average cost per word" c i s
l i nked t o our
def i ni t i on of t he energy
redundancy Rof t hesyst em
t hrought he
si mpl e rel at i onshi p
c a ( 1 - R), a
mi ni mi zat i on of c corresponds t o a
maxi -
mi zat i on of R.
Expressi on ( 2) can t heref ore
be consi dered as t he t heoret i cal
sol ut i on descri bi ng
t he f requency di st ri but i on
of a syst emof
words wi t h a
gi ven const ant
val ue I at equi l i bri um;
-t hat i s correspondi ng t o
t he maxi mum
val ue of R=
Rma=.
4.

THESECOND
LAWOFGENESI S
Mandel brot ' s
anal ysi s l i mi t s i t sel f t o
syst ems const i t ut ed of words ;
each
word bei ng an ensembl e of l et t ers separat ed
f romot her words
by space.
The Genesi s
Model -Part I I

( 51L) 107
We conj ecture that
the
arguments devel oped above can
be extended
to any
sel f organi zed system at
any
l evel of organi zati on andpostul ate a general l aw.
The
second
l awof
genesi s: The rank-ordered f requency di stri buti on of
subsystems/ el ements
of
any sel f organi zed system at
any
hi erarchi cal l evel of
organi zati on can be
approxi mated by a mathemati cal
di stri buti on
of the
f orm
p. = A( n +m)-B ( 2)
where A=

1/ E( n
+
m) - $, B=const . , n the rank-order andma parameter
n
i nf l uenci ng the di stri buti on f or smal l n onl y.
I n the f ol l owi ng we wi l l put f orward some evi dence i n f avour of thi s
hypothesi s.
5.

FREQUENCYDI STRI BUTI ONOFCHEMI CAL ELEMENTS
I NTHEUNI VERSE
I n the anal ysi s of secti on 3, we
consi dered
l anguage as a sel f organi zed
systemconsti tuted of el ements ( words) separatedby space. Def i ni ng mutual l y
excl usi ve types of el ements ( word-types) i n terms of the number of
components and thei r sequence at the next l ower hi erarchi cal l evel ( l etters of
an al phabet) we determi ned the rank-ordered f requency di stri buti on of
el ement-types, whi ch can general l y be approxi mated by a mathemati cal l aw
of f orm( 2) .
I n anal ogy we consi der the uni verse as asel f organi zedsystemconsti tuted
of el ements ( atomi c i sotopes) separated by space
.
Def i ni ng mutual l y excl usi ve
types of el ements ( chemi cal el ements) i n terms of
the number
of components
and thei r sequence at the next l ower hi erarchi cal l evel ( el ectrons, protons) we
determi ned the rank-ordered f requency di stri buti on of chemi cal el ement-
types. Fi gure 3
( on the f ol l owi ng page) based on the data compi l ed by J . P.
Meyer and A. G. W.
Cameron
as
quoted by H.
Reeves( ' ) i ndi cates
that thi s
di stri buti on
can
be
descri bed by a l awof type ( 2) . Except f or n <7( i nf l uence
of the parameter m), the empi ri cal data can surpri si ngl y wel l be approxi mated
by a strai ght l i ne.
6.

FREQUENCYDI STRI BUTI ONSI NBI OLOGI CAL SYSTEMS
( i ) Asi ngl e speci es consti tuted
of
ensembl es
of
i ndi vi dual s. Consi deri ng
a l ocal ani mal popul ati on
bel ongi ng to
a
si ngl e speci es
of
parasi tes andthei r
hosts as a sel f organi zed system consti tuted of el ements ( ensembl es of
i ndi vi dual s on si ngl e hosts) separated
by space,
we
def i ne mutual l y excl usi ve
types of el ements i n terms of thenumber of i ndi vi dual s
i n
an
ensembl e, that
i s
by the number of parasi tes on asi ngl e host .
Fi gure 4 ( on the f ol l owi ng pages) shows an
exampl e
of a rank-f requency
di stri buti on of parasi te ensembl es
based
on
the
data
quotedby C. B.
Wi l l i ams
( 7)
( i i ) Asi ngl e genus consti tuted
of
speci es. Consi deri ng a l ocal
i nsect
popul ati on
bel ongi ng
to a
si ngl e genus onl y as a sel f organi zed system
consti tuted of el ements ( speci es), we def i ne
mutual l y excl usi ve types of speci es
i n terms of the number of i ndi vi dual s consti tuti ng each speci es.
108( 5
2L)

Pet er
Wi ni wart er
10 100 1000
RANK-ORDERn -
Fi gure 3.

The rank-f requency di st ri but i on of chemi cal el ement s
i n t he uni verse .
Fi gure
5 ( opposi t e) shows a rank-f requency di st ri but i on
of speci es-t ypes
based on t he
dat a of C. B. Wi l l i ams
( 7),
whi ch are t he resul t
of dai l y random
sampl es of t he
genus Macrol epi dopt era caught
i n a l i ght t rap at Rot hamst ed
duri ng
f our successi ve years ( pat i ence agai n! ) . Al t oget her
t here were
15, 609
i ndi vi dual s
caught , represent i ng 240speci es.
( i i i ) A
si ngl e f ami l y const i t ut ed of genera.
I n anal ogy t o ( i i ) one can cl i mb
up t he hi erarchy and consi der
t he' ani mal popul at i on bel ongi ng t o a si ngl e
f ami l y as a sel f organi zed syst em
const i t ut ed of el ement s, ( genera) and
def i ne
mut ual l y excl usi ve genus-t ypes
i n t erms of t he number of speci es const i t ut i ng
each genus. Fi gure 6
( on t he f ol l owi ng pages) shows a rank-f requency
di st ri but i on of genus-t ypes
based on t he cl assi f i cat i on of
t he i nsect f ami l y
Mant i dae
by
W. F
. Ki rkby ( i n 1910) as quot ed by C. B. Wi l l i ams
( 7) .
TheGenesi s Model -Part I I

(53L)
109
RANK.-ORDER n
Fi gure
4.
Rank-f requency
di stri buti on of ensembl es of i ndi vi dual s bel ongi ng to one
bi ol ogi cal
speci es.
c
RANK-ORDER n -
1OO
Fi gure 5. Rank-f requency di stri buti on of speci es bel ongi ng to
one bi ol ogi cal genus.
110( 5
4L)

Pet er Wi ni wart er
RANK-ORDER n -
Fi gure 6.
Rank-f requency di s t ri but i on of genera
bel ongi ng t o one bi ol ogi cal f ami l y .
7 .

FREQUENCYDI STRI BUTI ONSI NECONOMI C
SYSTEMS
Cons i deri ng
a capi t al i s t economy as
a s el f organi zed s ys t em cons t i t ut ed
of
el ement s ( monet ary uni t s repres ent i ng
s ervi ces or goods , l i ke words
repres ent i ng
abs t ract not i ons or phys i cal obj ect s ) we
def i ne mut ual l y excl us i ve
t ypes of
el ement s ; each
el ement t ype bei ng charact eri zed by t he
ent erpri s e
produci ng i t .
Fi gure 7 ( oppos i t e) s hows
t he rank-f requency di s t ri but i on of
monet ary
uni t s produced
by t he 1000great es t French
bus i nes s ent erpri s es bas ed on t he
dat a of
t he year 1980 compi l ed by Dun &
Brads t reet as publ i s hed by
C. Bari onet t al
.
Except f or n <7 ( i nf l uence of t he
paramet er m) t he 1000empi ri cal dat a
poi nt s can be approxi mat ed by a s t rai ght
l i ne i n doubl e-l ogari t hmi c coordi nat es
wi t h
an accuracy rarel y f ound i n experi ment al
s ci ences .
8.

CONCLUSI ONOF PARTI I
Des pi t e t he s mal l s ampl e
of s ys t ems anal yzed,
t aki ng i nt o cons i derat i on
t hei r di vergent nat ure, we
can s ay t hat t he empi ri cal dat a
s peak i n f avour or
- gi ven t he rat her poor s t at i s t i cs
f or bi ol ogi cal s ys t ems - do
not obvi ous l y
cont radi ct t he hi ghl y s pecul at i ve
s econd l awof genes i s .
Af urt her prel i mi nary
s urvey concerni ng s el f -organi zed
s ys t ems of mat t er
has reveal ed t hef ol l owi ng res ul t :
The Genesi s Model -Part I I
the rank-f requency di stri buti ons
of chemi cal
el ements
i n
the
uni verse, of
masses i n the sol ar system( sun +pl anets) , of masses i n the Saturn system
( pl anet +satel l i tes) , of chemi cal el ements i n the earth shel l , of chemi cal
el ements i n sea-water
and
of
chemi cal
el ements i n the human
body can
al l be f ai rl y wel l descri bed by di stri buti ons of f orm( 2) ; andwhat seems
even
more
surpri si ng - the sl opes of the
strai ght
l i nes i n
bi l ogari thmi c
coordi nates, correspondi ng
to
the
constants B,
are very
si mi l ar and
coul d
reveal i denti cal wi thi n error l i mi ts .
on the
other hand the
sl opes of
the strai ght l i nes descri bi ng
the
rank-
f requency di stri buti on
of
ensembl es
of
i ndi vi dual s of
a bi ol ogi cal speci es
i n an eco-system, of words i n al i ngui sti c systemandof monetary uni ts
i n
an
economi c systemare al so very si mi l ar
and
coul d
be descri bedby a
si ngl e constant .
I n vi ew of these prel i mi nary resul ts, the hypothesi s put f orwardi n Part I ,
that one andthe same quanti tati ve pri nci pl e governs the process of evol uti on
at
al l l evel s
( , ~C >0)
seems to us deservi ng of f urther i nqui ry.
10
11
10
10
-_
n~
-
A( n i 1)
-
R
10

100

1000
RANK-ORDER n -+
( 55L) 111
Fi gure 7. Rank-f requency di stri buti on of monetary uni ts i n
a capi tal i st economy.
9. OUTLOOK
Part I I I wi l l present the actual genesi s model , whi ch has l ed to the
devel opment of our concept
of
a quanti tati ve measure of compl exi ty
.
I ntro-
duci ng the noti on of abstract
automata -
a
mathemati cal concept i ni ti al l y
deri ved
f rom the study of sequenti al swi tchi ng ci rcui ts - the process of
evol uti on coul dbe descri bedas thesel f organi sati on of oneabstract automaton
f ol l owi ng one and the same al gori thmthroughout i ts organi zati on i nto more
andmorecompl exsubsystems or abstract sub-automata.
112( 56L) Pet er
Wi ni wart er
Acknowl edgement s
I woul d l i ke t o express mygrat i t ude t o al l t he
sci ent i st s whose pat i ent
compi l at i on work has made t he
above
qui ck
surveypossi bl e.
Ref erences
1.

Wi ni wart er,
P. , Spec
. Sci . Tech. ,
6,
11 ( 1983) .
2.

Zi pf ,
G. K. , Human Behavi our and t he Pri nci pl e of Least Ef f ort , Addi son- Wesl ey,
Cambri dge,
Mass.
( 1949) .
3.

Shannon,
C. E. and Weaver,
W. ,
The Mat hemat i cal Theory of Communi cat i on, Uni -
versi t yof I l l i noi s Press, Urbana ( 1949)
.
4.

Cherry,
L. , On
Human Communi cat i on,
M. I . T. Press, Cambri dge, Mass. ( 1966) .
5.

Mandel brot , B. , Cont ri but i on a t o t heori e
mat hemat i que des j eux de communi cat i on,
Thesi s
Sc
. Mat h, Pari s
( 1952) .
No.
3393
publ i shedi n Ext r
.
des
Publ i cat i ons de l ' i nst i t ut
de St at i st i que de
1' uni versi t
de Pari s, 2, Ease. 1 and
2,
Pari s
( 1953) .
6.

Reeves, H. , Pat i ence dans 1' azur, L' evol ut i on cosmi que, Edi t i ons du Seui l , Pari s ( 1981).
7.

Barj onet , C. , L' Expansi on, 6, 109- 309( 1981) .

You might also like