Professional Documents
Culture Documents
FRANC1SCO J. VARELAt
School of Medicine, University of Cobrado, Denver, Colorado, U.S.A.
(Rccclvad December 6,1977; infinalform May 26. IPf8)
This paper p m t s a lrarncwotk within which a number or prcrcrnd syncm drxriptiont mn bc unitid. Or pnrticutar
interest are [he difkrcrms stemming ttom thc study ol natunl systems (b~ologlnland social) and man-mdde systems
~(enginaringand computer). It is claimed that the insighrs Itom [he nudy o l n3!urrl sptcms have becn dewlopcd leu
~xrens~vcly and prmscly. One important case discussed here is the relationship bctwcen dcxrib~nga system as
~vtoaomowor as cnntrof!d, k prcciw chancterizal~onof complcmunrar~tyIn d a a l p t ~ o n ris propused through the
notion or adjoint runnor. to cxplicatc the interplay of these dual views. Othcr descriptive dualitla of similar kind a r t
dzxurzcd; the cornplemeniarity h t w c c n nets and I ~ 15Sprcxnled in rull detail.
INDEX TERMS Complemmr~rity.distinaions nutonomy, control, adjoint functor, rrces. networks.
1. INTRODUCTION
. . diridinn the world into cnrironmeat and system,
in line- with their diverent purpcB, and have
The world does not present itself to us neatly also dtveloped different methodologics and termi-
divided into systems, subsystems, environments,
and so on. These are divisions which we make
nologies consisten! with their motivations.
In this paper, we prwent a framework within
ourselvts, for various purposes, often subsumed
which a number of these various preferred views
under the general purpase evoked by saying "for
convcnitnce". I t is cvtdent that different people
on systems can be unified. Ot particular interest
find it convcnicnt to divide the world in different
ro as are the differences stemming from the study
OF natural systems (particularly biological and
ways, and even ant person will be interested in
social systems) and man-made systems (such as
different systems at different time, for example,
engineering and computer systems).
now a cell, with. the rest or the world its environ-
Contemporaty system theory has developed ex-
ment. and later she postal system. or the econ-
tensively through experience in the Iattr fields,
omic system, or the atmospheric system.
but thc insights derived from natural sysrcms
The established scientific disciplines have, of have remained by and large much less formally
course, developed different preferred ways of developed. We hold that the notions of cooper-
ative interaction, self-organization, and
tnk papr wu Prtidbs u ~ ~ r t cby d the Narov sutonomy-in brier, holistic notions-a~ re-
Institulc b u ! d a , Colorado (J.G.). and rht A l i m l P, Slonn
Foundation (F+v,). is a stria or thrtc. -rhe wnd levant and basic to the study or systems.
paptr w ~ l lprtscnb some rnathcrnatical roundations lor sell- In the Framework of this Paper and its sumes5ors,
releren!~al g r m fin systems. including a Spacer- these nolions ate not only made more explicit
Brominn ~ l c u l r uof indiiuoad rszntry. third paw applicable, but arc also presented eomple-
will deal with applrcntions to variom spmfic uajcs.
:Prucnlt addfeu: Bmn Research h b o r a t o n q New Yore
ments to the more traditional notions of system
Uaivtr~ityMedical School. SSO Fint Avenue, New York. City theory, such as control and input-output h-
1W0016. havioral description.
Downloaded By: [UNAM] At: 23:17 12 October 2007
ver may choose to focus on the Internal structure largely ignores the cell. A similar hierarchy of
of the system viewing the environment as ba- levels can be found in thc social sciences. It seems
ckground, for example, as a sourcc or Ipertur- to' be a general reflection a[ the richness oh
bations upon the sy stem's autonomous behavior. natural systems that indimtion can be iterated to
From this viewpoint, the properties of the system product a hierarchy of levels.
emerge from the interactions d its compontn ts. At a given level of the hierarchy, a particular
Biology has iterated this proms of indication, system can be seen as an outside to systems
creating a hierarchy of levels or biological study. below it, and as an inside to systems above it;
The cell biologist emphasitts the cetl's autonomy, thus, the status (i.e., thc mark of distinction) of a
and views the organism of which it is part as given system changes as one passes through its
little more than a source o l perturbations for level, in either the upward or the downward
which the cell compensates. But the physiologist direction. The choice of considering the levcl
views the cell as an element in a network o l above or belaw corresponds 20 a choice of treat-
intcrdtpendencia constituting the individual or- ing the given system as autonomous or eon-
ganism: this corresponds to a wider view or trolled (constrained). Figures I to T illustrate a
environment, namely the ecology in which the variety of configurations aC systems. subsystems,
individual participates. A population biologist and marks. and Figure 8 illustrates the hierarchy
makes his distinctions at a still higher Icvel, and or levcls.
mation about the system's internal state, but the autonomy of the cell; if the bialogisr's prelertnm
emphasis ('mark") is on the environment, which for input and output variables do not match the
is identified with the observer. Behavior appears celI's internaI organization, the biologist's theory
M an input+utput function y(e.t), the observ- will not work. Furthemore, the hierarchy of
able results or applying the inputs (also called levels seem to particularCy assm its importance
Ccon~rols")e to the system. for natural systems, so that it is generally nec-
An alternative elaboration or the situation of essary to take account of. at least three levels.
(21 views the vector e as not ncmsarily or Even when the lowmt lcvel is very. well under-
particularly under the control of an observer, but stood, the role which it plays, at the next higher
rather as a source or perturbations upon (2). For level when interconnected with other systems, can
example, the components el of e may be some be qoite obscure. For example, an enzyme bio-
coeficients which art regarded as constants in chemist may be able to describe a particular
the original equation (1). A natural question to metabolic loop very ellecdvcly by a transfer Iwnc-
pose is the stability of the system under such tion, but Ibe quite unable to spbciFy haw it fits
perturbaaons, that is the relation of (1) to a into the overall metabolic process of the cell as a
penur'bed system coherent whole.
This situation or being unable to understand
haw elements, even quite well underslood elc-
mznts. coordinate or somehow runetion efhec-
in which 6 {in a rairly intuitive nonation) repre- tirely together at the next higher lcvel, is quite
sents a "small change". It is known, lor example, common in the study of narural syslems,'and is a
that changes in structural constants can cause the major source or our desire for a better developed
system to undergo a "catastrophic" change (in the theory of autonomous systems.
sense d ThornM)into a new configuration. Some Fragments of theory emphasizing the
The system ( 2 ) has in it nothing which intrinsi- aumnomy or systems d o exist, but they are far
cally prefers the approach of either (3) or (4). less developd than is the behavioristic approach.
This choice depends on rhe interest of the First and foremost, thc idea OF stability derived
analyst. [ o m classical mechanics, has been extensively
Note that "recursion" {in [he broad sense or studied and used. As wc said before, a set of
reedback which affects the hrrturc) plays a role in interdependent differentiaI equations can be used
all t h e e f~rmulations,but is more obscure in the to represent the autonomous properties oi a
control theory interpret ation, On the other hand, whole syslem. Rosen? Iberall'O and lange"^
the behavioral information, though still available, have applid this perspective lo natural systems
is more obscure In the stability interpretation (4). with various degrees or ernphnsis on autonomous
We are not, OF course, claiming that either or behavior. More speclific examples can be found in
these approaches is inhcrcntly better. population biology," cellular b i o l ~ p y , ' ~ -and
~'
Historically speaking, some of the many pos- more recently, in nturobiology." Some thought
sible approaches to systems hare k e n much has been given to cooperative interactions in this
more developed than others. The most highly area or hierarchical multilevel systems. The idea
developd parts, in fact, center on the notions of of hierarchy is of~enpresented from the p i n t or
control, input+utput behavior, and state- view of the interdependency or different levels of
transition, This is presumable because of the systems dcs~ripzions.'~."~'~Particular instances
interest in applying these approaches in of hierarchical structure including rnuItiIeveF
tnginmring. cooperation can be Found in "- 2 0 + 2 ' .G o g ~ t n ~ ~ . ~ ~
I t seems, however, that the notion of autonomy p m n ts a genera! theory of hierarchical systems
is panicularly important for natural systems, e.g., of interdependent processes. Its basic ideas are
biological and social systems, and the lack o l a interconnection, behavior, and lcvel, and its
well-developed theory of autonomous systems is theoretical framework is categorical algebra. A
felt as a serious difliculty. An engineer designing East area in which the idea of a whole system is
an artifact wilt choose the inpults of interest to somewhat explicit is that or self-organizing sys-
him tor this application with some assurance that tems. Work in this area based on an information-
the choice will be adequate. But a biologist theoretic. approach includes ron Foerster,'" and
studying a cell is Forced to acknowledge the A11an.I~
36 J. A. GOGUEN AND f. J. YARELA
Downloaded By: [UNAM] At: 23:17 12 October 2007
We do not intend to unite all these various The nodes 1,2,3 might represent three physical
rhretlds of research together in a single kame- locations, each with a CB (Citizens Band) radio,
work. Rather. we emphasize the ways in which Because of direrencts in transmitter and rmivzr
pairs of seemingly different points or view, such strength, available frequencies, and terrain, corn.
as autonomy/control, are cumplementary, in the munication is possible only along the channels
sense or conthibutiag ta a better understanding of indicated by arrows in (5). For example, there is
natural systems. But 'the idea of complementarity, a mountain bttwcen I and 4. k t us assume that
hundarnental though it seems, is still vague. The node 1 is of particular interest-say it is our
next two sections develop an txpiicit definition, base. Thcn we are interested in the patterns of
4. NETS AND TREES transmission which are possible starti& from our
base. For example, to reach node 3, if channel g
II we retain inttrtst only in the connectivity of a is out, we can send a message via f ij; to verify its
system, it is possible to represent the duality correctness, it couId k sent back to node 1, via
recursion/behavior by 'a networkJtret duality. the pathjijk; node 3 might also generate a reply,
Intuitively, the nodes in these nets or trees repre- which would require a message to be sent to
sent the elcrnents or components of a system, node 2, giving a path ijkj; and so on. Thus, we
while their links reprcscni interactions or in- arc interested in the set of all paths in G with
terconnections. Tht reciprocal connectivity of a source 1. This c~llection itself has a branching
net suggests the coordination of a system's ele- structure, because starting with a given path, i t
ments; a tree structure suggests the sequential can sometimes be developed rurther by choosing
subordination or a system's parts, each part hnv- alternative edgcs to get alternative paths. The
ing its own wel1delined input-output behavior collection of all choices can be represtnted by [he
description. To be sure, in retaining only the following tree
basic connectivity of a system's organization.
much is disctlrded in the netttree representation.
We intcnd to use this convenient gencrol repre-
sentation to study camplemenrurity.
Now ro the definition of nets and 'trees. Let
therc be a set of ( v , . .... w,} OF nodes (components
or parls), which arc to be interconnected by a set
or E = {c,. ....c . ~ edges ) (relations or processes).
1. DEFINITION A r~eswork is n directed graph
C, that is, a quadruple G=(IC[,E,do.2,), where
- J,:E+IGI
1Gb= { D , , , . . , u ~ } . and are the source (i
) and target (i = 1 3 functions, from the edges
to the nodes ot G. lf e ~ E , d , e = v and d , c = u',
Notice that it is an infinite tree,
then we write e : v + v ,
I n some sense the tree (6) "unravels'" or "un-
+
2. DEFINITIONA path from v to v' in a graph folds" the graph ( 5 ) rrom node 1. To make this
G is a finite sequence p = e, . ..e, 06 edges which more precise w t define tsm, pointcd graphs, and
arc adjacent, that is, sa~isfya, e, =Joe,, for I 2 i structure-preserving mappings ot graphs called
s n , with doeo= w and ?,en==". lr d,p=v and d,p graph homomorphism.
= v', then write p: v - . ~ ' ,
4. DEFINITION A poinrsd graph t is a 5-tupfc
3. EXAMPLE Consider the graph G: (IG[,E,~,.C,,~) such that <] G l . ~ , 2 , , d , > is a
1 graph and a ~GIl is a vertex.
k jlZ A pointed graph is reachable iT for each vertex
v s ] G I there is a path a+v in G.
1
A graph G is loopfree if for all u, Y'E GI there
is at mosl one path v-v'.
A tree is a reachable 1oopiree pointed graph,
5. DEFINITION Let G be a pointed graph
(1GI,E, Eo, S,, 0 ) . Then the unfoldmtnr U,(G) of G
Downloaded By: [UNAM] At: 23:17 12 October 2007
rrom a, is the graph in which: lCI,(G)I is all the in a similar way for the node pa-, , and the path
paths p:a-.o, for YEIG~;
the edges of U J G ) are p r 4 pm - , , p ,-,, and p ' k p ,-,, and so on.
the pairs (p,pe), such that p, p e ~ Iu,(E)(,
and Eventually we must find that p'=p, for some k,
e E E ;do{p, pe) = p, and a , { p , pe} = pe. The null and the unique path p'+p is of the form
path o+a is written l,:o-~u, and is taken to be.
the point For U , ( G ) ,
6. P~omsrnon Let C be o pointed graph
{I G 1, E, 3,. a,. a>, and U J C ) the unfildmnr of G If p=p', the unique path p+p' is the null path at
from a. Then U,(G) is a trre. p. Thus U,(G) is loopfree, and the proof, is
Proof We must show that U J G ) is a pointed, complete,
reachable, and loopfree graph. By the definition :
of CI'JG), it has point 1,: a+a. 7. DEF~NITION
Let G={lGl.E,a,,J,> and G'
We now show that U , ( C ) is reachable.
... =(IG'~, E;db,a',) be graphs. Then a raph mor-
Consider a node p :a+o of Wa(G), say p=e, e, phism is a pair ( J FI, F) a l luncrions I Ff : G -1 Cl
# id, with e, E E. Then we can show that: and F : E - c E ' , such that the source and targel
relationships are preserved, that is, such that
and
is a path horn 1, to p in U , ( G ) . Clearly its source
is 1, since
i.e. such that the diagram
general setting of category theory, which is be- The discussion at the end of the previous
coming increasingly useful in systems theory."*'' section shows that the netltrix complementarity
Readers unfamiliar with this terminology may is an instance of the concept of, adjunction. What
find a leisurely introduction in ' 0 - 3 ' or we we now propose, is to explore she view [hat rhe
attempt to stay at a fairly intuitive level in this precise concept of udjuncrion is on explication of
Section, although some technicalities are the general (and vogue) concept of
inevitable. complementarity.
The intuitive idea of a category is that it Another example a1 this is Goguen's adjun.
embodies some structure by txhi biting the class ction lxtween minimal realization and be-
of all objects having that structure, together with h a v i ~ r . * ~ ~Let
' ' A be the category o l automata
all the structure prwerving mappings or mor- (in some fixed smse which we shall not explain in
phisms among them. (Somewhat wore technically, detail), and let B k the category of input-output
categories assume there is an associative oper- behaviors of such automata (with appropriate
ation of composition on t hosc morphisrns whosc rnorphisms). Then there is a functor from B to A
source and target match.) This idea is due to which constructs the minimal au tomalion M ( B )
Eilenberg and M a c L a n ~ . ~ ~ having the behavior 3;and there is a functor Be
For example, pointed graphs and pointed from A to '& which constructs"the behavior Be(A )
graph morphisms constitute a category. If C 1s a of an automaton A. Moreover, thtre is a natural
category, and A , B are objects in C, we shall ler bijection
C ( A , B ) denote the set of all morphisms in C
rrom A to 8.
Usually, we are intcrestcd no[ only in objects
rrom various categories, but we are even more which expresses the complementarity of the no-
interested in certaln constructions performed on tions of internal stale transition (as embodied in
the objccts ol one category to yield objects of automata) and input-ouaput behavior. GoguenJ5
another category. For example, unfoldment is a has shown that many other classes or systems
construction performed upon graphs which yields exhibit such a complementarity with their input-
t recs. This construct ion has a kind of consistency, ourput behaviors.
in that i t can nlso be extended to the morphisrns; Hert is still another example. IT G is a graph,
that is, a rnorphism or pointed graphs induces in the collection of at1 paths (From all sourws) in G
a natural way, a morphism betwecn their unfold- forms a category whosc objects art the nodes of
ments. This kind or consistency is expressed by G, and whosc rnorphisrns are the paths of G;this
saying that the construction is funcrorial, or is a category is denoltd Pa(G), and called the path
Jmcrar. {More technically, this has to do with the caregory OF G. Pa is a functor from the category
preservation of the composition or rnorphisms). Cr of graphs to the category Cat of (small)
However, the unColdment construction is na- categories. There is also a functor F from Cot to
tural in n much stronger sense: the ''optimal" Gr, which merely forgets the additional structure
covering of a graph is its unioldrnent; this is which categories have over graphs (namely. the
expressed by the universal property of possibility of composing morphisms), regarding
Theorem 8, and the bijection q ol the previous the objects as nodes, and the morphisms as
section. The conccpk of.adjunction generalizes just edges. Again, there is a natural bijection
this state of affairs, to the lollowing situation:
Let A and B be categories, let F be a lunctor C)-.Gr (G, FC)
q : Cat (Pa(C),
from A to 0 , and C a functor from 0 to A. Then
an adjuncrion is, in addit ion, a natural bijection. expressing the complementarity of graphs and
categories. Alternatively, Pa(G) is the free ca-
tegory generated by the graph G, and the adjun-
This says that every rnorphism f : FA+B de- ction (or [he corresponding "universal property")
termines a unique rnorphism ~ ( f A-.GB.
) : (The expresses this relationship. (See3& for further
precise sense of the '"naturalness" or cp is that of details.)
n a t u a l transformation, due to Eilcnbtrg and In general, it is more satisfacltory to consider
M a ~ L a n c , ~which
' however, we shall not define diflcrtnt modes of dexription that may appear as
here. The idea o i adjunction is due to Kan.") opposites. to be complementary instead. This is
Downloaded By: [UNAM] At: 23:17 12 October 2007
4
l J. A. GOGUEN 41
the case, quite rigorously, with the apparent well find some application in general discussions
dualities ntt/trec and recursion/behavior. as we about system theory. Much more work, including
have seen above. On a more intuitive level, there many Further exampPes, wiIl be needed to dis-
is a similar relatianship for the pair auto- cover the proper domain of application, and the
nomy/control discussed in earlier sections. limits, of the adjointless idea.
LawvereJ4 is a particularly fundamental, stimu- The general system theory of G o g ~ e n ~in- ~*~'
lating paper, which suggesrs that there is.a cont- volves a hierarchy of Iewels, much as pictured in
plernen tary relationship between the traditional Figure 8, with functors going outwardJ9 which
conceptual/fomal viewpoints in the foundations regard a component at a lower 1cvcl as a whole
or mathematics. This duality also appears as a system at the next higher level, and runetors
semantics/syntax pair, in that LawvereJS has going inward which compute the behavior of the
shown an adjunction between a functor which whole system, viewing the result as a single
associates to each algebraic theory its category of object at the lower level. Thcre is a base level o l
semantic models (i.c., Its algebras), and a Functor glven 'objects" our of which systems can be
which extracts from each category, the optimal constructed, and objects at level ( i + l ) arc in-
syntactic t32tory or its algebraic component of terconncctions (that is systems) of objects at ltvel
structure. i. Goguen shows that each pair of outward/in-
VarelaJn has described many situations where ward functors is an adjunction. The inward iunc-
apparent opposites arc in Iact complementary. tor is in Iact the tundarntntal categorical con-
Out usual way of thinking about pairs of de- struction known as "limit". Goguen also shows
scriptive notions is biased in the direction of that the consrruction of interconnecting a system
regarding them us mutual negations, or the g- of systems (over some common subparts as "ter-
n e r d lorm: not-A/A. I t is the case, however, minals") 10 get a single systcm, is given by the
that many of the most important such pairs dual concept or "colimit", which also appears as
operatc by mutual specification. that is. are com- an adjunction. This is not the place to give
plemcntnry. We mention in particular, content/ details, bat the connecrion with the themes of
' Torm, scmantics,'syntax. ~utonomy/controt, net/ this paper should bc evident.
tree. minimal-model/behaviw, sirnultaneous/
sequential, and tnvironment/system. OC course,
the gcrminal example is the fernale/rnale com- 6. HOLISM AND REDTTCT10NISM
pkmcntnrity, which seems to bc particularly mis-
understood as nn opposition in our contem- If we think OF the philosophy of.science, the
porary culture. dual~ty holism/~ductionism comes to mind as
This general point seems particularly clear in analogous to the material previously discussed in
the context or systems theory: there is no whole this paper.
system without an interconnection or its pans: Most discussions place hoIism/reductionism in
and there is no whole system without an environ- polar opposition.a"."~41 This stems no stem from
ment. Such pairs are mutually interdependent: the historical split ktwetn empirical sciences,
tach defines the other. What is remarkable about viewed as mainly rductionist or analy~ic.and the
the notion or adjoint functor, is that it captures (European) schools of philosophy and social
the notion OF complementarity in a very precise science that grope toward a dynamies of
way, without imposing any particular model lor totaEiti~~.*~"*
the nature of the objects so related. It is also In the light of the previous discussion, both
worth noting that there is a well-developed attitudes are possible lor a given descriptive level,
theory of adjunctionsr for example. the corn- and in fact they arc complementary. On the onc
position OF ~ w o adjoint pairs of functors is an- hand, one can move down a level and study the
other adjoint pair. Such results should be very properties of the components. disresarding their
useful in system theory. Of course, not all pairs mutual interconnection as a system.On theother
of descriptive modes are eornpttmzntary, and hand, one can disregard the detailed structure of
similarly, not all pairs of runctors are adjoint. the components, treating their behavior only as
The so-called "adjoint runctor t hcorem" provides contributing to that or a larger unit. I t seems.thnt
some generill conditions For when a given functor both these directions of analysis always coexist.
in iact does have an adjoint, and again, this may either implicitly or explicitly, because these de-
Downloaded By: [UNAM] At: 23:17 12 October 2007
sc'riptivt levels are mutually interdependent Ior behavior of organisms to the behaviors of mo-
the observer. We cannot conceive of components Iecules, we may say that organisms are whole
if t h t r t is no system from which they are abstrac- systems. Similarly, it is very diflicult (if not
ted; and there cannot be a wholt unless there are impossible) to reduce the effects or melodies to
constitutive clemtnts. the effects of notes. Onc must consider properties
. It is interesting to consider whether one can ol patterns of notes or molecules.
have a measure &r the degree of wholeness of a A third point of.vitw is that a system is whole
system. One can, of course, always draw a distinc- to the extent that its parts art lightly intereon-
tion, make a mark, and get a "system" but the necred, that is, to the degree that ir is dificult to
rtsuh does not always seem to be equally a find relatively independent subsystems. This is
"whole system*',a "natural entity", or a "coherent clearly related to the previous views. An interest-
object" .or "concept". What is it that makes some ing corollary or this view is that a system with a
systems more cohtrtnt, more natural, more strongly hierarchical organization will be lws
whole, than others? (See45,7~a6, for different ways whole than a system with a strongly heterarchical
to answer thesc questions.) organization; that is, ntrs arc mort whole than
One thing ta notice is, that in the hierarchy or trees. More precisely, given that the graph OF
levels, "emergent" or "immanent" properties a g connections or the parts or a system has no
pear at some levels. For cxamplc, Ict us consider isolated subsystems, the more tree-like it IS, the
music as a system or organization of notes (for less whole it is, while still k i n g (presumably) a
the purpose or this example, we do not attempt system. The extreme is probably a pure linear
to reduce notes to any lower lcvel distinctions). structure, without any branching at all.
Then harmony only arises when we consider the A lounh point of view, is that a system seems
simultaneous or parallel sounding of notes, and more wholt if it i s more complex, that is, mort
melody only arises when we consider the sequtn- dificdt no reduce to descriptions as intercon-
tial sounding of notes. That is, harmony and nections of lower level components. I t is nec-
melody are emergent properties of a level of w a r y in this discussion to take account of the
organization above that OF the notes themset ves. very modern point or view that the relatively
, can only emerge at a still higher
~ i m i l a r l ~farm more complex a system is to describe. the more
level of organization. relating dimerent melodic random it is; in fact, K ~ l r n o g o r c vwho
~ ~ gave
units to one another, These properties, form, the original foundation for probability theory in
melody and harmony. arc systems properties, terms of measure theory," has proposed to redc-
arising from hierarchical organizations oC notes. fine probabiliky theory and informntion theory in
into pieces or rnusa; they are not properties of terms of complexity theory. Thus, for example,
notes, (See47 For further discussion of the hle- the wholeness or a living sptem is. in everyday
rarchical organization of music.) I t also appears encounters, construed as unprdictability. The
that "life" is an emergent property or the biologi- more difticult it is to reduce a system to a simple
cal hierarchy of levels: it is nowhere to be found inputJoutput control the more likely it is we will
at the lcvcl of aroms and rnoleculcs; but it deem it alive, In this sense complete autonomy is
becomes cleat at the level of cells through the logically cquivalenl 10 complete randomness.
autcpoietic organization or molecule^.'^-'^ Another example: a piece oi music which is too
Language can be been as an emergent property at comptex (rtfative to our cultural expectations and
a stilt higher level: of this hierarchy." inherent capacities) will sound random, chaotic,
Thus, one point of view toward wholeness is perhaps mcaninglcss, but it will also sound
that it co-o&uss with interesting emergent pco- whole. Here the extreme is white noise. (See4' lor
ptrtits at some level. A sequencc of notes is a discussion ob hierarchical complexity with a g
whole if it is an interesting melody. Similarly, the plications to musical atsthet icr.) fhis viewpoint
significant differences in behavior or cells and toward wholeness involves measurement relative
organism rrom the behaviors or atoms and mo- to some standard interpreting syslem, such as a
lecules, marks the cells and organisms as whole human being. But given such a standard, this
systems. viewpoint can be deduced lrom the preceding
Another point. OF view toward wholeness, is viewpoints. For surely, if it is dificult to describe
that i t can be measured by the diflculty of a system. it will also be diflicutt to reduce i t to
rerlrrcrion: Because it is very hard to reduce the lower ltvtls. and its parts w ~ l lseem to k tightly
Downloaded By: [UNAM] At: 23:17 12 October 2007
interconnected- Quite possibly, its very corn- already indicate a prdcrmct lor marking S; that h, tk
tanguagc incoqmtalcs the prcltrtna. But wt may s p k
ploxity will appear as an emergent property. o l 'marking tht cavironmnt" to suggest that thcrt are in
On the other hand, a diflerent cognitive view- k t two
~ distinct pssibilitin.
point might well 'be b e ~ t t rable to p r o w what 6. H von Focnter, 'Nola for an Epistemology or Living
now seems likc a very complex system, and thus Thine." CUnite dc L"h4mnu. cdittd by E. Morin and M.
see it as less whole. Once again, the relativity to du Stuil, Paris. 1974, pp.401417.
Piatelli E ~ L
7. F. Vattla md J, Gogum T h e krithmctit of Clasun." I.
cognitive rapacity appears, This can bc sem over Cybcncfiu. 8, 1978.
and over in the history of. science, as new tech- 8. R. Thom. Stabilifi Sfwturclk er Morphaginiw.
niquw, tools, or ideas, suddenly make long stand- Benjamin, N m York. 1972
ing problems soluble, or perhaps, uninteresting or 9. R. Rosm. bymmical S p t c n u Thcary in Biology. John
Willty, N w York. 1972
even ill-posed. 1Q. A. Ibcrdl, Towor& a GtncraI Oicnce Q! b b l e System
Thest descriptive levels haven't been gmeralIy McGraw HilL. Ncw Yorlc. 1972
realized as complementary largely because there 1 I. 6.Langc, Wholes and Parrr Pergamon P ~ e y New York.
is a diflerence between publicly announced 1965.
methodology nnd actual practice, in most fields 12 R. May, Mdcl Ecmyrremr. Princeton U. Prey
Prinfeton. 1971.
or research in modern science. A reductianist 13. M. Eigcn, %if+rganhtian and the Evolution of
attitude is strongly promoted, yet the analysis 06 Matter.* Narunviu. 5%, 1971, pp. 46S523.
n system cannot begin without acknowledging a 14. B. C.-win, Tcmporul Organizurion of CeIk A d c m i c
dcgrec of coherence in the system lo be in- Psty New York, 1968.
15. A. Katchalsky. V. Rowland and R. Blumenthal.
vestigated; the analyst has to have an intuition "Dynam~e Patterns or Brain Cell Auemblia."
that he is actually dealing with a coherent pheno- Ncuroscict~csRex. Prog, Bull., 13. NO.1. 1974.
menon. Although science has publicly taken a 16. H. Pattee, d, Hierarchy Theory. Gmrgc B&llier, Ntw
rductionisz attitude, in practice both approaches York. 1972.
have always been active. I t is not that one has to 11. L. Whytr. A. Wilson and D. Wilwn. Hicrnrrhical
Structures. Elscvicr. New York, 1968.
have a holistic view as opposed to a rcductionist IS. M . Maarov~c, D. Macka, Y. Takahark Tbavry of
view, or uice uuersa, but rather that the two views Hirrnrchtcal Multiltrel Sptcm. Academic P m , New
ofsystems arc cornpbmen~ary. Vork, 1972
There is a strong current in contcmprary 19. S. k r , The Brain oJ thc Plrm. Allm Lane. Candon. 1912.
20. L. Kohour and B. Gainer, -Prokc*lon a a &era1
culture advocating "holistic" views as some sort System Problem." Inr. J. Gen Sysrrmt. 3, 1976. pp, 3-24
o l cute-all. This is easily seen in discussions of 21. T.Baumgnrtner. Z: Burng L. Meeker and R. Wild, uO~pen
environmental phenomena, health delivery, and System and Multi-Level P r m : Impltations lor
value systems. The way in which holism is used h n l Racareh." Pnr. J. Gm. Systrmr. 3. 1976. pp. 2-2.
in this paper is quite diretmt. We simply see 22. J. A. Goguen. -Mathematical Rsprttcntarion or
Hierarchically Orgtnucd Sysltrns: Clobul Syftmm
reductionism and holism as complementary, or 5 ~ m m i c edited
~, by E . Attin~cr.S. Karger, B a d , 1971.
"cognitively adjoin t", Tor the dmript ions of those Pi. 112-128.
systems in which we ate interested. Reductionism 23. J. A. Gosum. 'Svstmu and Minimal R e a t i t i o n . *
implies attention to a lower level, while holism ~ r o f c c d i a ;I$-197; IEEE Conftrcncc on D ~ b l u ma d
Control. M i h i b c h 197L pp. 4 2 4 .
implies attention to a higher level. Thtse are 24. H,van Focrster, -On 511-Organizing S p t m and Their
intertwined in any satisfactory description; and Environments: Se!F4rgonizing Sptrrm, edited by M.
each entails some loss relative to our cognitive Yovits and S. Cameron, Pcrpmon Press. London. 1%6.
pretcrcnces, ns well as some gain, pp. 51-50.
55. H. Atlan. d0rgani:arion Btolo#tquc ei la Tharrrv dc la
Ir$ormtion Herman, Park 1972.
26. 1. A. Goguen. "On Homomorphisms. ~ o m t r ; c u ,
REFERENCES AND NOTES Termination. Unroldmtnts. and Equivnlenm of Flow
aagram Prosrams." Journal of Conputrr and Sprem
I. G. Spenmr-Brown.I m w ~of Form b n r a m Books. New Sciences, I, No. 3. 1974. pp. 533-365.
York. 1973. 27. Such a 'elass" is a son of higher logical type, consisting
2. F. Ynrsla. 'A Cnfculus Tor Sctf-RcTctcnce," In!. I. Gem. on nll xrs satisfying some conditions. Such a notion h u
Sy~rrmr,5. 1975. pp. 5-24. bten tormliad wtthin CWel-&mays set lhmry. and
3. G. Bataon, 'A theory of Plny and Fanray." Reprinted dacr not lead lo contradictions.
in: G. Bateson. Srrps Towards m Ecology oj Mind. 28. J. A. Goguen, "Realization IS Universal.' Mathcmtkaf
Bullankine. New York, 1972. System Theory, 6 , p10.4. 1973, pp. 359-374.
4. E. CotTman. Frume Anulysb. Harper, Colophon Bouks. 29. M , brbib and E. Mancr. 'Machines in a Category: An
New York, 1974. E x p i l o r y Introduct~on.' SIAM Rcv, 16. 1974. pp. 163-
5. Calling S *the $ygtemh rather than 'the snuironmmt" 191.
Downloaded By: [UNAM] At: 23:17 12 October 2007