VOL. 48, NO. 2 March, 1973 THE QUARTERLY REVIEW of BIOLOGY

LIVING SYSTEMS

BY JAMES G. MILLER

Departmentof Psychiatry and BehavioralSciences, School of Medicine, The JohnsHopkins University, Baltimore, Maryland*

ABSTRACT

General systemsbehavior theory is concerned with seven levels of living systems cell, organ, organism, group, organization, society, and supranational system. The following article is an exposition of the basic concepts in this integrativetheoretical approach. It is a condensation of a more detailed statement (Miller, 1965a; 1965b; 1965c). The second article is an analysis in terms of this conceptual system of present knowledge concerning one level of living system-the organism. In order to empha- size the cross-levelformal identitiesamong levels of living systems,a major consideration of general behavior systemstheory, this article follows exactly the same outline, with identical subheadings and section numbers, as other articles written by the author. These deal with the lower levels of living systems-cell and organ (Miller, 1971a)- as well as higher levels-group (Miller, 1971b), organization (Miller, 1972), society, and supranational system. (All the articles will be published togetheras chapters of the author's forthcomingbook, Living Systems.) The primary intent of this article, in addition to its analysis of the content it covers, is to show that the structuresand processes of organismsand of living systemsat all the levels are directlycomparable. Since anatomists and physiologists are usually laymen in organization theory or international relations, psychologistsare commonly laymen in economics, and social scientistsare ordinarily laymen in cellular biology, all parts of the book, including the two following articles, are necessarily writtenfor intelligent laymen rather than experts, even though the articles deal with many technical topics. Some statements in them will seem to experts to be too elementaryto be worth repeating. If a fact is fundamental and may not be known to specialists in other fieldsit is stated here, even if it is elementary to the experts. The complex division of labor of modern science, often characterized by pluralistic insularity,requires this. The multitude of detailed and specialized experimentsand studies that have been carried out provide the substance of the scientificinvestigation of organisms. Their findingsconstitute the trees. But an overview of these results and of the relationships among them- a view of the forest- is also essential. Such a telescopic rather than

* AfterSeptember 1, 1973,President, University of Louisville. 63

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 64 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48

a microscopicview may suggest the proper balance for research on various aspects of organismsand clarifythe priorities for future efforts. The following articles view organisms broadly. They deal with all living systemsat this level -protistans, fungi, and plants, as well as animals. Why such a range? To make manifestthe continuityof organismic life in structure,process, and history (including evolution). To demonstratehow structureand process go toget,her,so that anatomists, physiologists,and psychologistsunderstand the common task they share and how each field can enrich the others. Most of the ideas presented in these articles are not new even though the concep- tual integration is original with the author. The concepts are derived from many aspects of science and have been developed by many workers,including various former associates of the author at the Universities of Chicago and Michigan, and systems scientists in many other places. The articles necessarily select for discussion only a few researches out of the vast published repertoire,and this selection has necessarily been arbitrary. Experts in each special field might agree on other studies as more important. Some of the author's statemnentsmay be wrong and his analysis ill advised. If so he would appreciate corrections-it is hard to cover such a wide range and still make no errors. The constant referencein the second article to cross-levelhypotheses stated in the firstarticle has the purpose of showing that propositions possibly valid at other levels may also apply to organisms. At each level there are scientists who apply system theory in their investigations. They are systems theorists but not necessarily general systemistheorists. They are general systemstheorists only if they accept the more daring and controversialposition that -though every living system and every level is obviously unique -there are importantformal identitiesof large generalityacross levels. Such cross-levelsimilarities can potentially be evaluated quantitatively,applying the same model to data collected at two or more levels. This possibility is the chief reason why the author has used the same outline with identically numbered sections to analyze the present knowledge about each of the seven levels of living systems. The following survey of what is known about organisms as systems,therefore, is to be read as a single segment of a larger,integrated whole.

I. THE NATURE OF LIVING SYSTEMS G ENERAL SYSTEMS behaviortheory mathematicsmay have any number of dimen- is a set of related definitions,as- sions. sumptions,and propositionswhich Physicalspace is the extensionsurrounding a deal with reality as an integrated point. Classically the three-dimensionalgeom- hierarchyof organizationsof matter etry of Euclid was considered to describe ac- and energy. General systemsbehavior theory curately all regions in physical space. The is concernedwith a special subsetof all systems, modern general theoryof relativityhas shown the livingones. that physical space-timeis more accuratelyde- Even more basic to thispresentation than the scribedby a geometryof fourdimensions, three conceptof "system"are the conceptsof "space," of space and one of time. "time," "matter,""energy," and "information," This presentation of a general theory of because the living systemsdiscussed here exist living systemswill employ two sorts of spaces in space and are made of matter and energy in which theymay exist,physical or geograph- organizedby information. ical space and conceptual or abstractedspaces.

1. SPACE AND TIME 1.1 Physical or Geographical Space In the most general mathematical sense, a This will be considered as Euclidean space, space is a set of elements which conform to which is adequate for the study of all aspects certain postulates. The conceptual spaces of of living systemsas we now know them.Among

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973 3 THE NATURE OF LIVING SYSTEMS 65 the characteristicsand constraintsof physical section. They can give the location of objects space are the following: (a) From point A to in it. point B is the same distance as frompoint B to -point A. (b) Matter or energy moving on 1.2 Conceptual or AbstractedSpaces a straightor curved path frompoint A to point B must pass through every interveningpoint Scientificobservers often view living systems on the path. This is true also of markersbear- as existing in spaces which they conceptualize ing information. (c) In such space there is a or abstract from the phenomena with which maximum speed of movement for matter, they deal. Examples of such spaces are: (a) energy,and markersbearing information. (d) Peck order in birds or other animals. (b) Social Objects in such space exert gravitationalpull class space (lower lower, upper lower, lower on each other. (e) Solid objects movingin such middle, upper middle, lower upper, and upper space cannot pass through one another. (f) upper classes). (c) Social distance among ethnic Solid objects moving in such space are subject or racial groups. (d) Political distance among to frictionwhen they contact another object. political parties of the right and the left. (e) a scale of The characteristicsand constraintsof physical Sociometricspace, e.g., the rating on each member of a group space affectthe action of all concrete systems, leadership ability of (f) A space of time living and nonliving. The followingare some by every other member. of various modes of transportation,e.g., examples: (a) On the average, people interact costs longer on foot than by air, longer more with persons who live near to them in a travel taking than down. housing project than with persons who live upstream and abstracted spaces do far away in the project. (b) The diameter of These conceptual and are not the fuel supply lines laid down behind General not have the same characteristics Patton's advancing American Third Army in subject to the same constraintsas physicalspace. of its World War II determinedthe amount of fric- Each has characteristicsand constraints tion the lines exerted upon the fuel pumped own. These spaces may be either conceived of about from throughthem, and thereforethe rate at which by a human being or learned such fuel could flowthrough them to supplyPatton's others. Interpreting the meaning of and measuringdis- tanks. This was one physical constraintwhich spaces, observingrelations, limited the rate at which the army could ad- tances in them ordinarilyrequire human ob- servers. Consequently the biases of individual vance, because they had to halt when theyran human beings color these observations. out of fuel. (c) Today informationcan flow Social and some biological scientistsfind con- worldwidealmost instantlyby telegraph,radio, ceptual or abstractedspaces useful because they and television. In the SeventeenthCentury it that physical space is not a major took weeks for messages to cross an ocean. A recognize determinantof certain processes in the living governmentcould not send messagesso quickly systemsthey study. For example, no matter to its ambassadorsthen as it can now because where they enter the body, most of the iodine of the constraintson the rate of movementof atoms in the body accumulate in the thyroid the marker bearing the information. Conse- gland. The most frequent interpersonalrela- quently ambassadorsof that centuryhad much tions occur among persons of like interestsor more freedomof decision than they do now. like attitudesrather than among geographical Physical space is a common space, for the neighbors. Families frequentlycome together reason that it is the only space in which all for holidays no matter how far apart their concrete systems,living and nonliving, exist membersare. Allies like England and Australia (though some may exist in other spaces simul- are often more distant from each other in taneously).Physical space is shared by all scien- physicalspace than theyare fromtheir enemies. tificobservers, and all scientificdata must be It is desirable that scientistswho make ob- collected in it. This is equally true for natural servationsand measurementsin any space other science and behavioral science. Most people than physical space should attemptto indicate learn that physical space exists, which is not precisely what are the transformationsfrom true of many spaces I shall mentionin the next their space to physical space. Other spaces are

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 66 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48 definitelyuseful to science, but physical space of state of matter-energyor its movementover is the only commonspace in which all concrete space, fromone point to another,is action. It systemsexist. is one formof process.

1.3 Time 3. INFORMATION This is the fundamental"fourth dimension" Throughout this presentation information of the physical space-time continuum. Time (H) will be used in the technical sense first is the particular instant at which a structure suggestedby Hartley (1928). Later it was de- exists or a process occurs, or the measured or veloped by Shannon (1948) in his mathematical measurable period over which a structureen- theoryof communication. It is not the same dures or a process continues. For the study thingas meaning or quite the same as informa- of all aspects of living systemsas we know tion as we usually understand it. Meaning is them,for the measurementof durations,speeds, the significance of information to a system rates, and accelerations, the usual absolute which processes it: it constitutesa change in scales of time- seconds,minutes, days, years that system'sprocesses elicited by the informa- are adequate. A concrete systemcan move in tion, often resultingfrom associations made to any direction on the spatial dimensions, but it on previous experience with it. Information only forward - never backward - on the tem- is a simpler concept: the degrees of freedom poral dimension. that exist in a given situationto choose among signals, symbols,messages, or patterns to be transmitted. The total of all these possible 2. MATTER AND ENERGY categories(the alphabet) is called the ensemble. Matter is anythingwhich has mass (m) and The amount of informationis measuredby the occupies physical space. Energy (E) is defined binary digit, or bit of information. It is the in physicsas the ability to do work. The prin- amount of informationwhich relieves the un- ciple of the conservationof energystates that certaintywhen the outcome of a situationwith energycan be neither created nor destroyed two equally likelyalternatives is known.Legend in the universe,but it may be convertedfrom says the American Revolution was begun by a one form to another, including the energy signal to Paul Revere fromOld North Church equivalent of rest-mass.Matter may have (a) steeple. It could have been either one or two kinetic energy,when it is moving and exerts lights"one if by land or two if by sea." If the' a force on other matter; (b) potential energy, alternativeswere equally probable, the signal because of its position in a gravitationalfield; conveyedonly one bit of information,resolving or (c) rest-massenergy, which is the energythat the uncertaintyin a binary choice. But it would be released if mass were convertedinto carried a vast amount of meaning, meaning energy. Mass and energyare equivalent. One which must be measuredby other sortsof units can be convertedinto the other in accordance than bits. with the relation that rest-massenergy is equal The term markerrefers to those observable to the mass times the square of the velocityof bundles, units, or changes of matter-energy light. Because of the known relationship be- whose patterningbears or conveysthe informa- tween matter and energy, throughout this tional symbolsfrom the ensemble or repertoire chapter the joint term matter-energyis used (von Neumann, 1958,p. 6-7) (1). These might except where one or the other is specifically be the stones of Hammurabi's day which bore intended. Living systemsrequire matter-energy, cuneiformwriting, parchments, writing paper, needing specific types of it, in adequate Indians' smokesignals, a door key with notches, amounts. Heat, light,water, minerals, vitamins, punched cards, paper or magnetic tape, a com- foods, fuels, and raw materials of various puter's magnetized ferrite core memory, an kinds, for instance, may be required. Energy arrangementof nucleotidesin a DNA molecule, for the processes of living systemsis derived the molecularstructure of a hormone,pulses on from the breakdown of molecules (and, in a a telegraph wire, or waves emanating from a few recent cases, of atoms as well). Any change radio station. The marker may be static, as

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 67 in a book or in a computer'smemory. Com- ode ray tube and photographed. The pattern munication of any sort,however, requires that of the chest structures,the information,modi- the markermove in space, fromthe transmitting fied for easier interpretation,has remained systemto the receivingsystem, and this move- largelyinvariant throughoutall this processing ment follows the same physical laws as the from one sort of marker to another. Similar movementof any other sort of matter-energy. transformationsgo on in living systems. The advance of communicationtechnology over One basic reason why communicationis of the years has been in the directionof decreas- fundamentalimportance is that informational ing the matter-energycosts of storing and patterns can be processed over space and the transmittingthe markerswhich bear informa- local matter-energyat the receivingpoint can tion. The efficiencyof informationprocessing be organized to conform to, or comply with, can be increased by lessening the mass of the this information.As already stated, if the in- markers,making them smaller so they can be formation is conveyed on a relatively small, stored more compactly and transmittedmore light, and compact marker,little energy is re- rapidly and cheaply. Over the centuries engi- quired for thisprocess. Thus it is a much more neering progress has altered the mode in efficientway to accomplish the result than to markersfrom stones bearing cuneiformto mag- move the entire amount of matter-energy, netic tape bearing electrons,and clearly some organized as desired, from the location of the limit is being approached. transmitterto that of the receiver. This is the In recent years systemstheorists have been secretof success of the deliveryof "flowersby fascinatedby the new ways to study and mea- telegraph." It takes much less time and human sure informationflows, but matter-energyflows effortto send a telegram from one city to are equally important. Systems theory deals another requesting a floristin the latter place both with informationtheory and with ener- to deliver flowers locally, than it would to getics- such matters as the muscular move- drive or flywith the flowersfrom the former ments of animals, the flow of raw materials city to the latter. through societies, or the use of energy by Shannon (1948, p. 380-382) was concerned neurons. with mathematical statementsdescribing the It was noted above that the movement of transmissionof informationin the form of matter-energyover space, action, is one form signals or messagesfrom a sender to a receiver of process. Another form of process is infor- over a channel such as a telephone wire or a mation processingor communication,which is radio band. These channels always contain a the change of informationfrom one state to certain amount of noise. In order to convey a another or its movement from one point to message,signals in channels must be patterned another over space. Communications, while and must stand out recognizably above the being processed, are often shifted from one backgroundnoise. matter-energystate to another,from one sort Matter-energyand informationalways flow of marker to another. If the form or pattern together. Information is always borne on a of the signal remainsrelatively constant during marker. Converselythere is no regular move- these changes,the informationis not lost. For ment in a systemunless there is a difference instance,it is now possible to take a chest x- in potential between two points,which is nega- ray, storing the informationon photographic tive entropyor information.Which aspect of film; then a photoscanner can pass over the the transmissionis most important depends filmline by line, fromtop to bottom,convert- upon how it is handled by the receiver. If the ing the signals to pulses in an electricalcurrent receiver responds primarilyto the material or which represent bits; then those bits can be energic aspect, it is a matter-energytransmis- storedin the core memoryof a computer;then sion; if the response is primarilyto the infor- those bits can be processedby the computerso mation, it is an informationtransmission. For that contrastsin the picture pattern can be example, the banana eaten by a monkey is a systematicallyincreased; then the resultant nonrandom arrangementof specificmolecules, altered patterns can be displayed on a cath- and thus has its informationalaspect, but its

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 68 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48 use to the mionkeyis chieflyto increase the (commonly verbs and their modifiers),or by energy available to him. So it is an energy logical or mathematical symbols, including transmission. The energic character of the those in computer simulations and programs, signal light that tells him to depress the lever which representoperations, e.g., inclusion, ex- which will give him a banana is less important clusion,identity, implication, equivalence, addi- than the fact that the light is part of a non- tion, subtraction,multiplication, or division. random, patternedorganization which conveys The language, symbols,or computerprograms informationto him. So it is an information are all conceptsand alwaysexist in one or more transmission.Moreover, just as living systems concrete systems,living or nonliving, like a must have specific forms of matter-energy,so scientist,a textbook,or a computer. theymust have specificpatterns of information. For example, some species of animals do not 4.2 Concrete System develop normallyunless theyhave appropriate informationinputs in infancy. As Harlow and A concretesystem is a nonrandomaccumula- Harlow (1962) showed, for instance, monkeys tion of matter-energy,in a region in physical cannot make proper social adjustment unless space-time,which is organized into interacting they interact with other monkeys during a interrelatedsubsystems or components. period between the third and sixth monthsof 4.2.1 Units. The units (subsystems,compo- theirlives. nents, parts, or members)of these systemsare also concrete systems(Hall and Fagan, 1956, p. 18). 4. SYSTEM 4.2.2 Relationships. Relationships in concrete The termsystem has a number of meanings. systemsare of various sorts,including spatial, There are systemsof numbersand of equations, temporal,spatiotemporal, and causal. systemsof value and of thoughit,systems of law, Both units and relationshipsin concrete sys- solar systems,organic systems,management sys- tems are empirically determinable by some tems,command and control systems,electronic operation carried out by an observer. In theo- systems,even the Union Pacific Railroad sys- retical verbal statements about concrete sys- tem. The meanings of "system"are often con- tems, nouns, pronouns, and their modifiers fused. The most general, however, is: A typicallyrefer to concrete systems,subsystems, systemis a set of interactingunits with rela- or components; verbs and their modifiers tionships among them (2). The word "set" usually referto the relationshipsamong them. implies that the units have some commonprop- There are numerous examples, however, in erties,which is essentialif theyare to interactor which this usage is reversedand nouns referto have relationships. The state of each unit is patterns of relationshipsor processes,such as constrainedby, conditioned by, or dependent "nerve impulse," "reflex,""action," "vote," or on the state of other units (3). The units are "annexation." coupled. 4.2.3 Open System. Most concretesystems have boun(laries which are at least partiallypermea- ble, permittingsizeable magnitudesof at least 4.1 Conceptual System certain sorts of matter-energyor information 4.1.1 Units. Units of a conceptual systemare transmissionsto cross them. Such a systemis terms,such as words (commonly nouns, pro- an open system.Such inputs can repair system nouns, and their modifiers),numbers, or other components that break down and replace symbols,including those in computer simula- energythat is used up. tions and programs. 4.2.4 Closed system. A concrete systemwith 4.1.2 Relationships. A relationship of a con- impermeable boundaries through which no ceptual systemis a set of pairs of units, each matter-energyor informationtransmissions of pair being ordered in a similar way. For ex- any sortcan occur is a closed system.No actual ample, the set of all pairs consisting of a concretesystem is completelyclosed, so concrete number and its cube is the cubing relation- systemsare either relativelyopen or relatively ship. Relationships are expressed by words closed. Whatever matter-energyhappens to be

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 69

within the systemis all there is going to be. the earth, produces stressesto which they can- The energygradually is used up and the matter not adjust. Under such stresses they cannot gradually becomes disorganized. A body in a survive. hermeticallysealed casket, for instance,slowly crumblesand its component molecules become 4.3 AbstractedSystenm intermingled.Separate layers of liquid or gas in a container move toward random distribu- 4.3.1 Units. The units of abstracted systems tion. Gravity may prevent entirely random are relationshipsabstracted or selected by an arrangement. observerin the light of his interests,theoretical 4.2.5 Nonliving system. Every concrete system viewpoint,or philosophicalbias. Some relation- which does not have the characteristicsof a ships may be empiricallydeterminable by some living systemis a nonliving system. operation carried out by the observer, but 4.2.6 Living systems.The living systemsare a others are not, being only his concepts. special subset of the set of all possible concrete 4.3.2. Relationships. The relationships men- systems,composed of the plants and the ani- tioned above are observed to inhere and in- mals. They all have the followingcharacteristics: teract in concrete,usually living, systems. In (a) They are open systems. a sense, then, these concrete systemsare the (b) They use inputs of foods or fuels to re- relationshipsof abstractedsystems. The verbal store their own energyand repair breakdowns usages of theoretical statements concerning in their own organized structure. abstractedsystems are oftenthe reverseof those (c) They have more than a certain minimum concerning concrete systems: the nouns and degree of complexity. their modifierstypically refer to relationships and the verbs and their modifiers(including (d) They contain genetic material composed predicates) to the concrete systemsin which of deoxyribonucleicacid (DNA), presumably these relationshipsinhere and interact. These descended fromsome primordialDNA common concrete systemsare empiricallydeterminable to all life, or have a charter,or both. One by some operation carried out by the observer. or both of these is the template-the original A theoretical statementoriented to concrete "blueprint" or "program"-of their structure systemstypically would say "Lincoln was Presi- and process from the moment of their origin. dent," but one oriented to abstractedsystems, (e) They are largelycomposed of protoplasm concentratingon relationshipsor roles, would including proteins and other characteristic very likely be phrased "The Presidency was organic compounds. occupied by Lincoln" (4). (f) They have a decider, the essential critical An abstractedsystem differs from an abstrac- subsystemwhich controls the entire system, tion, which is a concept (like those that make causing its subsystems and components to up conceptual systems)representing a class of interact. phenomena all of which are consideredto have (g) They also have certain other specific some similar "class characteristic."The mem- critical subsystemsor they have symbioticor bers of such a class are not thoughtto interact parasiticrelationships with other living or non- or be interrelated,as are the relationshipsin living systemswhich carryout the processesof an abstracted system. any such subsystemthey lack. Abstractedsystems are much more common (h) Their subsystemsare integratedtogether in social science theorythan in natural science. to form actively self-regulating,developing, Parsons-and Shils (1951) have attempted to reproducingunitary systems, with purposes and develop general behavior theory using ab- goals. stracted systems. To some a social systemis (i) They can exist only in a certain environ- something concrete in space-time,observable ment. Any change in theirenvironment of such and presumablymeasurable by techniques like variables as temperature,air pressure,hydra- those of natural science. To Parsons and Shils tion, oxygen content of the atmosphere, or the systemis abstractedfrom this, being the set intensity of radiation, outside a relatively of relationshipswhich are the formof organi- narrow range which occurs on the surface of zation. To themthe importantunits are classes

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 70 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48 of input-output relationships of subsystems systems. It is of paramount importance for rather than the subsystemsthemselves. scientiststo distinguishclearly between them. To use both kinds of systemsin theoryleads 4.4 Abstractedvs. ConcreteSystems to unnecessaryproblems. It would be best if one type of systemor the other were generally One fundamental distinction between ab- used in all disciplines. stractedand concretesystems is that the bound- All three meanings of "system" are useful aries of abstractedsystems may at timesbe con- in science,but confusionresults when they are ceptually established at regions which cut not differentiated.A scientificendeavor may through the units and relationships in the appropriatelybegin with a conceptual system physicalspace occupied by concretesystems, but and evaluate it by collectingdata on a concrete the boundaries of these latter systems are or on an abstractedsystem, or it may equally alwaysset at regionswhich include withinthem well firstcollect the data and then determine all the units and internal relationshipsof each what conceptual systemit fits.Throughout this system. paper the single word "system,"for brevity, will A science of abstracted systemscertainly is alwaysmean "concretesystem." The othersorts possible and under some conditions may be of systemswill alwaysbe explicitlydistinguished useful. When Euclid was developing geometry, as either "conceptual system" or "abstracted with its practical applications to the arrange- system." ment of Egyptianreal estate,it is probable that the solid lines in his figureswere originally 5. sTRucTuRE conceivedto representthe bordersof land areas or objects. Sometimes,as in Fig. I-1, he would The structureof a systemis the arrangement use dotted "constructionlines" to help concep- of its subsystemsand components in three- tualize a geometricproof. The dotted line did dimensional space at a given moment of time. not correspondto any actual border in space, This always changes over time (5). It may Triangle ABD would be shown to be con- remain relativelyfixed for a long period or it gruent to Triangle CBD, and thereforethe may change frommoment to moment,depend- angle BAD was equal to the angle BCD. After ing upon the characteristicsof the process in the proofwas completed,the dotted line might the system. This process, halted at any given well be erased, since it did not correspondto moment,as when motion is frozenby a high- anything real and was useful only for the speed photograph,reveals the three-dimensional proof. Such construction lines, representing spatial arrangementof the system'scomponents relationships among real lines, were used in as of that instant. the creation of early forms of abstracted systems. 6. PROCESS If the diverse fieldsof science are to be uni- fied, it would help if all disciplines were All change over time of matter-energyor in- oriented either to concrete or to abstracted formationin a systemis process. If the equa- tion describinga processis the same no matter whether the temporal variable is positive or negative, it is a reversible process; otherwise it is irreversible.Process includes the on-going a FIG. - A EUCIDEAFIGR_ functionof system,reversible actions succeed- ing each otherfrom moment to moment.Process also includes history, less readily reversed changes like mutations,birth, growth, develop- F_c 1- A EULDA FIGURE||. |- ment, aging, and death; changes which com- monly follow trauma or disease; and the changes resulting from learning which is not later forgotten.Historical processesalter both the structureand the function of the system.

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 19731 THE NATURE OF LIVING SYSTEMS 71

The statement"less readily reversed"has been 8. LEVEL used instead of "irreversible"(although many such changes are in fact irreversible)because The universecontains a hierarchyof systems, structuralchanges sometimescan be reversed: each higher level of systernbeing composed of a component which has developed and func- systemsof lower levels (6). Atomsare composed tioned may atrophyand finallydisappear with of particles; molecules, of atoms; crystalsand disuse; a functioningpart may be chopped organelles,of molecules. About at the level of offa hydra and regrow. History,then, is more crystallizingviruses, like the tobacco mosaic than the passage of time. It involves also ac- virus,the subsetof livingsystems begins. Viruses cumulation in the systemof residues or effects are necessarilyparasitic on cells, so cells are of past events (structuralchanges, memories, the lowest level of living systems. Cells are and learned habits). A living systemcarries its composed of atoms, molecules, and multi- historywith it in the formof altered structure, molecular organelles; organs are composed of and consequentlyof altered functionalso. So cells aggregated into tissues; organisms, of thereis a circularrelation among the threepri- organs; groups (e.g., herds, flocks, families, mary aspects of systems- structure changes teams, tribes), of organisms; organizations,of momentarilywith functioning,but when such groups (and sometimessingle individual orga- change is so great that it is essentiallyirreversi- nisms); societies,of organizations,groups, and ble, a historicalprocess has occurred,giving rise organisms;and supranational systems,of socie- to a new structure. ties and organizations.Higher levels of systems may be of mixed composition,living and non- living. They include planets, solar systems, 7. TYPE galaxies, and so forth. It is beyond the scope of this article to deal with the characteristics- If a number of individual living systemsare whatever they may be -of systemsbelow and observed to have similar characteristics,they above those levels which include the various often are classed togetheras a type. Types are forms of life, although others have done so abstractions. Nature presents an apparently (Neyman and Scott, 1957; Neyman, Scott, and endlessvariety of living thingswhich man, from Shane, 1954-1955). The subsetof living systems his earliest days, has observed and classified- includes cells,organs, organisms, groups, organi- first,probably, on the basis of their threat to zations, societies, and supranational systems. him, their susceptibilityto capture, or their It would be convenient for theoristsif the edibility,but eventuallyaccording to categories hierarchiallevels of living systemsfitted neatly which are scientificallymore useful. Classifica- into each other like Chinese boxes. The facts tion by species is applied to organisms,plants are more complicated. No one can argue that or animals, or to free-livingcells, because of there are exactly these seven levels, no more their obvious relationships by reproduction. and no less.For These systemsare classifiedtogether by taxono- example,one mightconceivably mistson the basis of likeness of structureand separate tissue and organ into two separate process,genetic similarityand ability to inter- levels. Or one might maintain that the organ breed, and local interaction,often including, in is not a level, since no organ exists that can animals,ability to respondappropriately to each live independentof other organs. other's signs. What are the criteriafor distinguishingany There are various types of systemsat other one level from the others? They are derived levels of the hierarchyof living systemsbesides from a long scientifictradition of empirical the cell and organism levels, each classed ac- observation of the entire gamut of living sys- cordingto differentstructural and processtaxo- tems. This extensive experience of the com- nomic differentia. There are, for instance, munity of scientificobservers has led to a primitive societies, agricultural societies, and consensus that there are certain fundamental industrial societies. There are epithelial cells, formsof organizationof living matter-energy. fibrotblasts,red blood cells, and white blood Indeed the classical division of subject-matter cells, as well as free-livingcells. among the various disciplines of the life or

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 72 THE QUARTERLY REVIEW OF BIOLOGY EVOLUME 48 behavioral sciences is implicitly or explicitly monly accepted only since Darwin. It is the based upon this consensus. justificationfor the labors of the white rat in It is importantto follow one proceduralrule the cause of man's understandingof himself. in systemstheory, in order to avoid confusion. Rats and cats, cats and chimpanzees,chimpan- Every discussionshould begin with an identifi- zees and human beings are similarin structure, cation of the level of reference,and the dis- as comparative anatomistsknow, and in func- course should not change to anotherlevel with- tion, as comparativephysiologists and psychol- out a specificstatement that this is occurring ogists demonstrate. (7). Systemsat the indicated level are called The amount of variance among species is systems. Those at the level above are supra- greater than among individuals within a systems,and at the next higher level, supra- species. If the learning behavior of cat Felix suprasystems.Below the level of referenceare is compared with that of mouse Mickey, we subsystems,and below themsubsubsystems. For would expect not only the sort of individual example, if one is studyinga cell, its organelles differenceswhich are found between Mickey are the subsystems,and the tissue or organ is and Minnie Mouse, but also greater species its suprasystem,unless it is a free-livingcell differences. Cross-species generalizations are whose suprasystemincludes otherliving systems common,and many have good scientificaccept- with which it interacts(8). ability, but in making them interindividual and interspecies differencesmust be kept in 8.1 IntersystemGeneralization mind. The learningrate of men is not identical to that of white rats, and no man learns at A fundamental procedure in science is to exactlythe same rate as any other. make generalizations from one system to The third type of scientificgeneralization anotheron the basis of some similaritybetween indicated in Fig. 1-2 is from one level to the systems,which the observersees and which another. The basis for such generalization is permits him to class them together. For ex- the assumption that each of the levels of life, ample, since the NineteenthCentury, the field fromcell to supranational system,is composed of "individual differences"has been expanded, of systemsof the previous lower level. These followingthe traditionof scientistslike Galton cross-levelgeneralizations will, ordinarily,have in anthropometryand Binet in psychometrics. greatervariance than the other sortsof general- In Fig. 1-2,states of separate specificindividual izations,since theyinclude varianceamong types systemson a specificstructural or processvaria- and among individuals. But theycan be made, ble are representedby 1, to I"* For differences and theycan have great conceptual significance. among such individuals to be observed and That thereare importantuniformities, which measured,of course,a variable common to the can be generalized about, across all levels of type, along which there are individual varia- living systemsis not surprising. All are com- tions, must be recognized (T1). Physiologyde- posed of comparable carbon-hydrogen-nitrogen pends heavily,for instance,upon the fact that constituents,most importantlya score of amino individuals of the type (or species) of living acids organized into similarproteins, which are organismscalled cats are fundamentallyalike, produced in nature only in living systems.All even though minor variations from one indi- are equipped to live in a water-oxygenworld vidual to the next are well recognized. rather than, for example, on the methane and Scientistsmay also generalize fromone type ammonia planets so dear to science fiction.Also to another (T1 to Tn). An example is cross- they are all adapted only to environmentsin species generalization, which has been com- which the physical variables, like temperature, hydration, pressure, and radiation, remain within relatively narrow ranges (Henderson, 1T*--L...T 1958). Moreover, they all presumably have arisen fromthe same primordialgenes or tem- ...ILi Ln plate, diversifiedby evolutionarychange. Per- haps the most convincing argument for the FIG. I-2. INDIVIDUAL, TYPE, LEVEL. plausibilityof cross-levelgeneralization derives

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 73 fromanalysis of this evolutionarydevelopment described by the same formal model. Con- of living systems.Although increasinglycom- versely,it may perhaps be shown as a general plex typesof living systemshave evolved at a principle that subsystems with comparable given level, followedby higherlevels with even structures but quite differentprocesses may greatercomplexity, certain basic necessitiesdid have quantitative similarities as well. not change. All these systems,if they were to survive in their environment,had, by some 8.2 Emergents means or other,to carryout the same vital sub- systemprocesses. At any given stage of evolu- The more complex systemsat higher levels tion, if a mutationhad eliminated any of these manifestcharacteristics, more than the sum of processes,the systemcould not have survived. the characteristicsof the units, not observed Consequentlyall livingsystems at all levels have at lower levels. These characteristicshave been the same criticalset of vital subsystemprocesses. called "emergents." Significantaspects of liv- While free-livingcells, like protozoans, carry ing systemsat higherlevels will be neglectedif these out with relative simplicity,the corre- they are described only in terms and dimen- sponding processes are more complex in mul- sions used for their lower-levelsubsystems and ticellular organisms like mammals, and even components. more complex at higher levels. The same A clear-cutillustration of emergentscan be processesare "shredded out" to multiple com- found in a comparisonof three electronicsys- ponents in a more complex system, by tems. One of these-a wire connecting the the sort of division of labor which Par- poles of a battery- can only conductelectricity, kinson has made famous as a law (Parkinson, which heats the wire. Add several tubes, con- 1957, p. 2-13). This resultsin formalidentities densers,resistors, and controls,and the new sys- across levels of systems,more complex sub- tem can become a radio, capable of receiving systemsat higher levels carryingout the same sound messages. Add dozens of other compo- fundamentalprocesses as simplersubsystems at nents,including a picturetube and severalmore lower levels. controls,and the systembecomes a television A formalidentity among concretesystems is set whichcan receivesound and a picture. And demonstrated by a procedure composed of this is not just more of the same. The third three logically independent steps: (a) recogniz- systemhas emergentcapabilities the second sys- ing an aspect of two or more systemswhich has tem did not have, emergentfrom its special de- comparable statusin thosesystems; (b) hypothe- sign of much greater complexity,just as the sizing a quantitative identity between them; second has capabilitiesthe firstlacked. But there and (c) demonstratingthat identitywithin a there is nothing mystical about the colored certain range of error by collectingdata on a merry-go-roundand racing childrenon the TV similar aspect of each of the two or more sys- screen-it is the output of a systemwhich can tems being compared. It may be possible to be completelyexplained by a complicated set formulatesome useful generalizations which of differentialequations such as electricalengi- apply to all living systemsat all levels. A com- neers write, including terms representingthe parison of systemsis complete only when state- characteristicsof each of the set's components. ments of their formal identities are associated with specificstatements of theirinterlevel, inter- 9. ECHELON type, and interindividual disidentities. The confirmationof formalidentities and disidenti- This concept may seem superficiallysimilar ties is done by research. to the concept of level, but is distinctlydiffer- What makes interindividual, intertype,or ent. Many complex living systems,at various interlevelformal identities among systemsim- levels, are organized into two or more echelons portant and of absorbing interest,is that- if (in the militarysense of a step in the "chain they can be conclusivelydemonstrated - very of command," not in the other militarysense differentstructures, which carry out similar of arrangementof troops in rows in physical processes,may well turn out to carryout acts space). In living systemswith echelons the com- so much alike that they can be quite precisely ponents of the decider, the decision-making

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 74 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48 subsystem,are hierarchicallyarranged so that systemis differentiatedfrom the environment. usually certain types of decisions are made by The immediateenvironment is the suprasystem one component of that subsystemand others minus the systemitself. The entire environ- by another. Each is an echelon. All echelons ment includes this plus the suprasuprasystem are within the boundaryof the decider subsys- and the systemsat all higher levels which con- tem. Ordinarily each echelon is made up of tain it. In order to survive the systemmust components of the same level as those which interactwith and adjust to its environment,the make up every other echelon in that system. other parts of the suprasystem.These processes Characteristicallythe decider componentat one alter both the systemand its environment.Liv- echelon gets information from a source or ing systemsadapt to their environment,and sources which process information primarily in returnmold it. The resultis that,after some or exclusivelyto and fromthat echelon. It may period of interaction,each in some sense be- be that at some levels of living systems-e.g., comesa mirrorof the other. cells-there are no cases in which the decidef is organized in echelon structure. 10.2 Territory Aftera decision is made at one echelon on The region of physical space occupied by a the basis of the informationreceived, it is trans- living system,and frequentlyprotected by it mitted, often through a single subcomponent froman invader, is its territory(Ardrey, 1966). which may or may not be the same as the de- Examples are a bowerbird'sstage, a dog's yard, cider,but possiblythrough more than one sub- a family'sproperty, a nation's land. component,upward to the next higherechelon, which goes througha similar process, and so 11. SUBSYSTEM AND COMPONENT on to the top echelon. Here a final decision is made and then command informationis trans- In everysystem it is possible to identifyone mitted downward to lower echelons. Charac- sort of unit, each of which carries out a dis- teristicallyinformation is abstracted or made tinct and separate process,and another sort of more general as it proceeds upward fromeche- unit, each of which is a discrete, separate lon to echelon and it is made more specificor structure.The totalityof all the structuresin detailed as it proceeds downward. If a given a systemwhich carryout a particular process componentdoes not decide but only passes on is a subsystem.A subsystem,thus, is identified information,it is not functioningas an eche- by the process it carries out. It exists in one lon. In some cases of decentralized decision- or more identifiable structural units of the making, certain types of decisions are made system. These specific, local, distinguishable at lower echelons and not transmittedto higher structuralunits are called componentsor mem- echelons in any form,while informationrele- bersor parts. Referencehas been made to these vant to other types of decisions is transmitted components in the definition of a concrete upward. If thereare multiple parallel deciders, systemas "a nonrandomaccumulation of matter- without a hierarchythat has subordinate and energy,in a regionin physicalspace-time, which superordinatedeciders, there is not one system is organized into interacting,interrelated sub- but multipleones. systemsor components." There is no one-to- one relationshipbetween processand structure. One or more processesmay be carried out by 1O. SUPRASYSTEM one or more components. Every systemis a 10.1 Suprasystemand Environment component,but not necessarilya subsystemof its suprasystem.Every component that has its The suprasystemof any living systemis the own decider is a systemat the next lower level, next higher systemin which it is a component but many subsystemsare not systemsat the or subsystem.For example, the suprasystemof next lower level, being dispersed to several a cell or tissue is the organ it is in; the supra- components. systemof an organism is the group it is in The concept of subsystemprocess is related at the time. Presumably every system has a to the concept of role used in social science suprasystemexcept the "universe." The supra- (Levinson, 1959). Organization theoryusually

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 75 emphasizes the functionalrequirements of the One can be betterthan another in maximizing systemwhich the subsystemfulfills, rather than effectivenessand minimizing costs. Valuable the specificcharacteristics of the component or studies can be made at each level on optimal componentsthat make up the subsystem.The patternsof allocation of processesto structures. typical view is that an organization specifies In all probability there are general systems clearly defined roles (or component processes) principles which are relevant to such matters. and human beings "fill them" (Weber, 1947). Possible examples are: (a) Structures which But it is a mistake not to recognize that char- minimizethe distanceover which matter-energy acteristicsof the component-in this case the must be transportedor informationtransmitted person carryingout the role-also influence are the most efficient.(b) If multiple com- what occurs. A role is more than simple "social ponents carryout a process,the processis more position,"a position in some social space which difficultto control and less efficientthan if a is "occupied." It involves interaction,adjust- single component does it. (c) If one or more ments between the component and the system. componentswhich carryout a process are out- It is a multiple concept, referringto the de- side the system,the process is more difficultto mands upon the component by the system,to integratethan if theyare all in the system.(d) the internal adjustment processes of the com- Or if thereare duplicate componentscapable of ponent,and to how the componentfunctions in performingthe same process,the systemis less meetingthe system'srequirements. The adjust- vulnerable to stressand thereforeis more likely ments it makes are frequentlycompromises be- to survivelonger, because if one componentis tween the requirementsof the component and inactivated,the other can carryout the process the requirementsof the system. alone. The way living systemsdevelop does not always result in a neat distributionof exactly 11.1 CriticalSubsystem one subsystemto each component. The natural arrangementwould appear to be for a system Certain processes are necessary for life and to depend on one structurefor one process.But must be carried out by all living systemsthat there is not always such a one-to-onerelation- surviveor must be performedfor them by some ship. Sometimesthe boundaries of a subsystem and a component exactly overlap, are congru- TABLE I-1 are not congruent. There ent. Sometimesthey THE CRITICAL SUBSYSTEMS can be (a) a single subsystemin a single com- ponent; (b) multiple subsystemsin a single SUBSYSTEMS component; (c) a single subsystemin multiple MATTER-ENERGY-WHICH PROCESS INFORMATION- components; or (d) multiple subsystemsin PROCESSING BOTHMATTER- PROCESSING multiplecomponents. SUBSYSTEMS ENERGYAND SUBSYSTEMS INFORMATION Systemsdiffer markedly from level to level, type to type,and perhaps somewhateven from Reproducer individual to individual, in their patterns of Boundary allocation of various subsystemprocesses to Ingestor Input Transducer differentstructures. Such a process may be (a) Internal Transducer localized in a single component; (b) combined Distributor Channel and Net with othersin a single component;(c) dispersed Converter Decoder laterally to other components in the system; Producer Associator (d) dispersed upwardly to the suprasystemor Matter-Energy Memory above; (e) dispersed downwardly to subsub- Storage systemsor below; or (f) dispersedoutwardly to Decider other systemsexternal to its hierarchy.Which Encoder allocation patternis employedis a fundamental Extruder aspect of any given system. For a specificsub- Motor Output Transducer functionin a specificsystem one strategy system Supporter results in more efficientprocess than another.

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 76 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48 other system.They are carried out by the fol- Supporter,the subsystemwhich maintains the lowing criticalsubsystems listed in Table I-1. proper spatial relationshipsamong components The definitionsof the criticalsubsystems are of the system,so that theycan interactwithout as follows: weightingeach other down or crowding each other. 11.1.1 SubsystemsWhich Process Both Matter- 11.1.3 Information-ProcessingSubsystems Energyand Information Reproducer, the subsystemwhich is capable Input transducer,the sensorysubsystem which of giving rise to other systemssimilar to the brings markers bearing informationinto the one it is in. (In Table I-1 a line is placed under system,changing them to other matter-energy it because it is the only subsystemcritical for formssuitable fortransmission within it. survival of the species but not critical for the Internal transducer,the sensory subsystem survivalof the individual system.A systemcan which receives,from all subsystemsor compo- survivewithout a reproducer,but not without nents within the system,markers bearing infor- any of the other subsystems.) mation about significantalterations in those Boundary, the subsystemat the perimeterof subsystemsor components,changing them to a systemthat holds together the components other matter-energyforms of a sort which can which make up the system,protects them from be transmittedwithin it. environmentalstresses, and excludes or permits Channel and net, the subsystemcomposed of entry to various sorts of matter-energyand a single route in physical space, or multiple information. interconnectedroutes, by which markersbear- 11.1.2 Matter-Energy-ProcessingSubsystems ing informationare transmittedto all parts of Ingestor,the subsystemwhich brings matter- the system. energy across the system boundary from the Decoder, the subsystemwhich alters the code environment. of informationtransmitted to it through the Distributor,the subsystemwhich carries in- input transduceror the internaltransducer into puts from outside the systemor outputs from a "private" code that can be used internallyby its subsystemsaround the systemto each com- the system. ponent. (It is usefulto distinguishthree sorts of codes, Converter,the subsystemwhich changes cer- increasing in complexity,which are used in tain inputs to the systeminto formsmore useful decoding and encoding- alpha, beta, and for the special processes of that particular gamma codes. An alpha code is one in which system. the ensembleof markersis composedof different Producer, the subsystemwhich formsstable spatial patterns of structuralarrangement of associations that endure for significantperiods physicalartifacts, like chemical molecules. Such among matter-energyinputs to the systemor arrangementsrepresent the coded "message" outputs fromits converter,the materials syn- or "signal" in a hormone or in such molecules thesized being for growth,damage repair, or as DNA and RNA. A beta code is one based replacement of components of the system,or on variationsin process,such as differenttem- forproviding energy for moving or constituting poral patternsof signals or differentpatterns the system'soutputs of productsor information of intensityof signals. Such codes are used by markersto its suprasystem. living systemswhich have decoders and en- Matter-energystorage, the subsystemwhich coders with stable programmingthat changes a retains in the system,for differentperiods of signal in one input code into a differentoutput time,deposits of various sortsof matter-energy. code. Neurons do this. A gamma code is a Extruder, the subsystem which transmits symboliclanguage used by systemswhich have matter-energyout of the systemin the formsof decoder and encoder subsystemsthat alter the productsand wastes. code on the input markersto a differentone on Motor, the subsystemwhich moves the sys- the output markersby comparingthe input to tem or parts of it in relation to part or all of a stored thesaurusof informationand selecting its environmentor moves components of its the output from it. This sort of symbolicin- environmentin relation to each other. formationtransmission is ordinarilydealt with

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 77

when thoughtand linguisticcommunication are 11.2 Inclusion discussed in the social sciences.) Associator, the subsystemwhich carries out Sometimesa part of the environmentis sur- and totallyincluded within the firststage of the learning process,forming rounded by a system Any such thingnot a part of the enduring associationsamong items of informa- its boundary. tion in the system. system'sown living structureis an inclusion. Any living systemat any level may include Memory,the subsystemwhich carriesout the living or nonliving components. The amoeba, second stage of the learning process, storing forexample, ingestsboth inorganicand organic various sorts of informationin the systemfor matterand may retain particles of iron or dye differentperiods of time. in its cytoplasm for many hours. A surgeon Decider, the executive subsystemwhich re- may replace an arterioscleroticaorta with a ceives informationinputs from all other sub- plastic one and that patient may live com- systemsand transmitsto them informationout- fortablywith it for years. To the two-member puts that control the entire system. group of one dog and one cat an important Encoder, the subsystemwhich alters the code plant componentis oftenadded - one tree. An of informationinputs to it fromother informa- airline firmmay have as an integralcomponent tion-processingsubsystems, from a "private" a computerizedmechanical systemfor making a code used internally by the system into reservationswhich extends into all its offices.A "public" code which can be interpreted by nation includes many sorts of vegetables,min- other systemsin its environment. erals, buildings, and machines, as well as its Output transducer,the subsystemwhich puts land. information from the out markers bearing The inclusion is a component or subsystem changing markers within the system system, of the systemif it carriesout or helps in carry- forms which can be into other matter-energy ing out a critical process of the system;other- over channels in the system'sen- transmitted wise it is part of the environment.Either way vironment. the system,to survive,must adjust to the in- Of these critical subsystemsonly the decider clusion's characteristics.If it is harmless or is essential in the sense that a systemcannot inert,it can oftenbe leftundisturbed. But if it be dependent on another systemfor its decid- is potentiallyharmful - like a pathogenic bac- ing. It can be dependent on another system terium in a dog or a Greek in the giant gift for any other critical subsystemprocess, but a horse within the gates of Troy-it must be living systemdoes not exist if the decider is rendered harmless or walled off or extruded dispersedupwardly, downwardly, or outwardly. from the systemor killed. Because it moves Since all livingsystems are geneticallyrelated, with the systemin a way the rest of the en- have similar constituents,live in closely com- vironment does not, it constitutes a special parable environments, and process matter- problem. Being inside the systemit may be energyand information,it is not surprisingthat a more serious or more immediate stressthan they should have comparable subsystemsand it would be outside the system's protective relationshipsamong them. All systemsdo not boundary. But also, the systemthat surrounds have all possible kinds of subsystems. They it can controlits physicalactions and all routes differ individually, among types, and across of access to it. For thisreason internationallaw levels, as to which subsystemsthey have and has developed the concept of extraterritoriality the structuresof those subsystems.But all living to provide freedom of action to ambassadors systemseither have their own full complement and embassies, nations' inclusions within for- of the criticalsubsystems carrying out the func- eign countries. tionsessential to life (i.e., theyare totipotential) or they lack some critical subsystems(i.e., are 11.3 Artifact partipotential) but are intimately associated with and effectivelyinteracting with systems An artifactis an inclusion in some system, which carryout the missing life functionsfor made by animals or man. Spider webs, bird them. nests, beaver dams, houses, books, machines,

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 78 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 4 music, paintings, and language are artifacts. nonliving artifactslike stones,books, and mag- They may or may not be prostheses,inventions netic tapes. It is not of itselfa concretesystem. which carryout some criticalprocess essential It changes only when man changes it. As long to a living system.An artificialpacemaker for as it is used it is in flux, because it must re- a human heart is an example of an artifact main compatible with the ever-changingliving which can replace a pathological processwith systemsthat use it. But the change emanates a healthy one. Insulin and thyroxinare re- from the users, and without their impact the placement drugs which are human artifacts. language is inert. The artifactual language Chemical, meclhanical,or electronic artifacts used in any information transmissionin a have been constructedwhich carry out some system determines many essential aspects of functionsof all levels of living systems. that system's structure and process (Whorf, Living systemscreate and live among their 1956). artifacts. Beginning presumablywith the hut and the arrowhead,the pot and the vase, the 12. TRANSMISSIONS IN CONCRETE SYSTEMS plow and the wheel, mankind has constructed All processinvolves some sort of transmission tools and devised machines. The Industrial among subsystemswithin a system,or among Revolution of the NineteenthCentury, capped There are across the boundary by the recent harnessingof atomic energy,rep- systems. inputs into a system,internal within it, and resents the extension of man's matter-energy processes outputs from it. Each of these sorts of trans- processingability, his muscles.A new Industrial missions may consist of either (a) some par- Revolution, of even greater potential, is just ticular formof matter; (b) energy,in the form beginning in the Twentieth Century,with the of light, radiant energy, heat, or chemical developmentof information-and logic-process- energy; or (c) some particular pattern of in- ing machines, adjuncts to man's brain. These formation. artifactsare increasinglybecoming prostheses, relied on to carryout critical subsystemproc- esses. A chimpanzeemay extend his reach with 13. STEADY STATE a stick; a man may extend his cognitiveskills When opposing variables in a systemare in with a computer. Today's prosthesesinclude balance, that systemis in equilibrium with re- input transducerswhich sense the typeof blood gard to them. The equilibrium may be static cells that pass before them or identifymissiles and unchangingor it may be maintainedin the that approach a nation's shores; photographic, midst of dynamic change. Since living systems mechanical,and electronicmemories which can are open systems,with continually altering store masses of informationover time; com- fluxes of matter-energyand information,many puters which can solve problems, carry out of their equilibria are dynamic and are often logical and mathematical calculations, make referredto as flux equilibria or steady states. decisions, and control other machines; electric These may be unstable, in which a slight dis- typewriters,high speed printers,cathode ray turbance elicits progressivechange from the tubes, and photographicequipment which can equilibrium states-like a ball standingon an output information. An analysis of many invertedbowl; or stable, in which a slight dis- modern systemsmust take into account the turbance is counteractedso as to restore the novel problems which arise at man-machine previousstate - like a ball in a cup; or neutral, interfaces. in which a slight disturbancemakes a change, Music is a special sort of human artifact,an but without cumulative effectsof any sort- information-processingartifact (Meyer, 1957; like a ball on a flat surfacewith friction. Cohen, 1962). So are the otherarts and cognitive All living systemstend to maintain steady systemswhich people share. So is language. states (or homeostasis)of many variables,keep- Whether it be a natural language or the ma- ing an orderlybalance among subsystemswhich chine language of some computersystem, it is processmatter-energy or information.Not only essential to information processing. Often are subsystemsusually kept in equilibrium,but stored only in human brains and expressed systemsalso ordinarilymaintain steady states only by human lips, it can also be recordedon with their environments and suprasystems,

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 79 which have outputs to the systemsand inputs times fit each other. That is, outside stresses from them. This prevents variations in the or threatsare mirroredby inside strains.Matter- environmentfrom destroying the systems.The energystorage and memoryalso mirrorthe past variables of living systemsare constantlyfluc- environment,but with certain alterations. tuating, however. A moderate change in one 13.1.1 Matter-EnergyStress. There are various variable may produce greater or lesser altera- ways for systemsto be stressed. One class of tions in other related ones. These alterations stressesis the matter-energystresses, including: may or may not be reversible. (a) matter-energyinput lack or underload- starvationor inadequate fuel input; (b) input an excess or overload of matter-energy;and 13.1 Stress,Strain, and Threat of (c) restraintof the system,binding it physically. There is a range of stabilityfor each of nu- [This may be the equivalent of (a) or (b).] merous variables in all living systems. It is 13.1.2 InformationStress. Also there are infor- that range within which the rate of correction mation stresses,including: (a) informationin- of deviations is minimal or zero, and beyond put lack or underload, resultingfrom a dearth which correctionoccurs. An input or output of informationin the environmentor from of either matter-energyor information,which improperfunction of the external sense organs by lack or excess of some characteristicforces or input transducers;(b) injection of noise into the variables beyond the range of stability, the system,which has an effectof information constitutesstress and produces a strain (or cut-off,much like the previous stress; (c) in- strains)within the system.Input lack and out- formationinput excess or overload. Informa- put excess both produce the same strain- tional stressesmay involve changes in the rate diminished amounts in the system. Input ex- of informationinput or in its meaning. cess and output lack both produce the oppo- site strain-increased amounts within the sys- 13.2 AdjustmentProcesses tem. Strains may or may not be capable of being reduced, depending upon their intensity Those processes of subsystemswhich main- and the resourcesof the system.The totalityof tain steady states in systems,keeping variables the strains within a systemresulting from its within their ranges of stabilitydespite stresses, template program and from variations in the are adjustment processes. In some systemsa inputs from its environmentcan be referred single variable may be influencedby multiple to as its values. The relative urgencyof reduc- adjustmentprocesses. As Ashby (1960, p. 153- ing each of these specificstrains representsits 158, p. 210-211) has pointed out, a living sys- hierarchyof values. tem's adjustmentprocesses are so coupled that Stressmay be anticipated. Informationthat the system is ultrastable. This characteristic a stressis imminentconstitutes a threatto the can be illustratedby the example of an army system.A threatcan create a strain. Recogni- cot. It is made of wires,each of which would tion of the meaning of the informationof such break under a 300-pound weight, yet it can a threat must be based on previously stored easily support a sleeper of that weight. The (usually learned) informationabout such situa- weight is applied to certain wires, and as it tions. A pattern of input information is a becomes greater, firstnearby links and then threatwhen -like the odor of the hunter on those fartherand fartheraway, take up part of the wind, a change in the acidity of fluids the load. Thus a heavy weight which would around a cell, or a whirlingcloud approaching break any of the component wires alone can the city-it is capable of eliciting processes be sustained. In a living system,if one com- which can counteract the stress it presages. ponent cannot handle a stress,more and more Processes -actions or communications-occur others are recruited to help. Eventually the in systemsonly when a stressor a threat has entire capacity of the systemmay be involved created a strainwhich pushes a variable beyond in coping with the situation. its range of stability. A systemis a constantly 13.2.1 Feedback. The term feedback (Ashby, changing cameo and its environmentis a simi- 1940; Rosenblueth,Wiener, and Bigelow, 1943, larly changing intaglio, and the two at all p. 19) means that there exist two channels,

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 80 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48 carrying information,such that Channel B extent that the systemdoes not survive. Feed- loops back from the output to the input of back control always exhibits some oscillation Channel A and transmitssome portion of the and always has some lag. When the organism signals emitted by Channel A (see Fig. 1-3.) maintainsits balance in space, thislag is caused These are tell-talesor monitorsof the outputs by the slownessof transmissionsin the nervous of Channel A. The transmitteron Channel A system,but is only of the order of hundredths is a device with two inputs, formallyrepre- of seconds. An organization, like a corpora- sented by a function with two independent tion, may take hours to correcta breakdownin variables, one the signal to be transmittedon an assemblyline, days or weeks to correcta bad Channel A and the other a previously trans- managementdecision. In a societythe lag can mittedsignal fed back on Channel B. The new sometimesbe so great that, in effect,it comes signal transmittedon Channel A is selected to too late. General staffsoften plan for the last decrease the strain resultingfrom any erroror war ratherthan the next. Governmentsreceive deviation in the feedback signal from a cri- rather slow officialfeedbacks from the society terion or comparison referencesignal indicat- at periodic elections. They can, however,get ing the state of the output of Channel A which faster feedbacks from the press, other mass the systemseeks to maintain steady. This pro- media, picketers, or demonstrators. Public vides control of the output of Channel A on opinion surveyscan accelerate the social feed- the basis of actual rather than expected per- back process. The speed and accuracyof feed- formance. backs have much to do with the effectivenessof When the signals are fed back over the feed- the adjustmentprocesses they mobilize. back channel in such a manner that they in- 13.2.2 Power. In relation to energy-processing, crease the deviationof the output froma steady power is the rate at which work is performed, state, positive feedback exists. When the sig- work being calculated as the productof a force nals are reversed, so that they decrease the and the distance throughwhich it acts. The deviation of the output froma steady state, it term also has another quite differentmeaning. is negative feedback. Positive feedback alters In relation to information-processing,power is variables and destroystheir steady states. Thus control, the ability of one "master" systemto it can initiate systemchanges. Unless limited, elicit compliance from another at the same or it can alter variables enough to destroysystems. a differentlevel. A systemtransmits a com- At everylevel of living systemsnumerous varia- mand signal or message to a given address bles are kept in a steady state, within a range with a signatureidentifying the transmitteras of stability, by negative feedback controls. a legitimate source of command information. When these fail, the structureand process of The message is often in the imperativemood, the system alter markedly-perhaps to the specifyingan action the receiver is expected

.~ ~FG I-3 NEGAIVEFEEDACK

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 81 to carryout. It elicits compliance at the lower maintaining adequate energylevels and there- levels because the electrical or chemical form fore it has the goal of ingestinga bacterium; of the signal sets off a specific reaction. At a boy has the purpose of keeping his body higher levels the receiving systemis likely to temperaturein the proper range and so he has comply because it has learned that the trans- the goal of findingand puttingon his sweater; mitteris capable of evokingrewards or punish- Poland had the purpose in March 1939 of re- mentsfrom the suprasystem,depending on how maining uninvaded and autonomous and so the receiverresponds. she sought the goal of a political alliance with 13.2.3 Purpose and Goal. By the information Britain and France in order to have assistance input of its charter or genetic input, or by in keeping both Germany and Russia from changes in behavior brought about by rewards crossingher borders. and punishmentsfrom its suprasystem,a system 13.2.4 Costsand Efficiency.All adjustmentproc- develops a preferentialhierarchy of values that esses have their costs, in energy of nonliving gives rise to decision rules which determineits or living systems,in material resources,in in- preferencefor one internal steady state value formation(including in social systemsa special rather than another. This is its purpose. It form of information often conveyed on a is the comparison value which it matches to markerof metal or paper money), or in time informationreceived by negative feedback in required for an action. Any of these may be order to determine whether the variable is scarce. (Time is a scarcityfor mortal living being maintained at the appropriate steady systems.)Any of these is valued if it is essen- state value. In this sense it is normative. The tial for reducing strains. The costs of adjust- systemthen takes one alternativeaction rather ment processesdiffer from one to another and than another because it appears most likely to from time to time. They may be immediate maintain the steady state. When disturbed, or delayed, short-termor long-term. this state is restoredby the systemby successive How successfullysystems accomplish their approximations,in order to relieve the strain purposes can be determinedif those purposes of the disparityrecognized internallybetween are known. A system'sefficiency, then, can be the feedbacksignal and the comparisonsignal. determined as the ratio of the success of its Any systemmay have multiple purposes simul- performanceto the costs involved. A system taneously. constantlymakes economic decisions directed A systemmay also have an external goal, toward increasing its efficiencyby improving such as reachinga targetin space, or developing performanceand decreasing costs. How effi- a relationship with any other system in the ciently a systemadjusts to its environmentis environment.Or it may have several goals at determined by what strategiesit employs in the same time. Justas thereis no question that selectingadjustment processes and whetherthey a guided missile is zeroing in on a target,so satisfactorilyreduce strains without being too there is no question that a rat in a maze is costly. This decision process can be analyzed searching for the goal of food at its end or by game theory,a mathematical approach to that the Greek people under Alexander the economic decisions. This is a general theory Great were seekingthe goal of world conquest. concerning the best strategies for weighing As Ashby (1961, p. 6-7) notes, natural selec- "plays" against "pay-offs,"for selectingactions tion permits only those systemsto continue which will increase profits while decreasing which have goals that enable them to survive losses,increase rewards while decreasingpunish- in theirparticular environments. The external ments, improve adjustments of variables to goal may change constantly,as when a hunter appropriate steady state values, or attain goals chases a moving fox or a man searches for a while diminishingcosts. Relevant information wife by dating one girl afteranother, while the available to the decider can improve such de- internal purpose remains the same. cisions. Consequently such information is A system'shierarchy of values determinesits valuable. But there are costs to obtaining such purposes as well as its goals. It is not difficult information. A mathematical theory on how to distinguishpurposes fromgoals, as the terms to calculate the value of relevant information have been used: an amoeba has the purpose of in such decisions was developed by Hurley

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 82 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48

(1963). This depends on such considerationsas ture at which process is most efficientis main- whetherit is tactical (about a specificact) or tained by a living system. strategic(about a policy for action); whether Hypothesis 3.2.2.2-1. The farthera specific it is reliable or unreliable, overtlyor secretly matter-energytransmission passes along a dis- obtained, accurate, distorted,or erroneous. tributorfrom the point of its input to it and towardthe finalpoint of its output fromit, the more it is alteredby loweringthe concentration 14. HYPOTHESES of the kinds of materialsor energyit contains which are used by the system'ssubsystems and A large number of general hypotheseswhich increasingthe concentrationof the productsor apply to two or more levels can be stated about wastes produced from it and output by those structureand process of living systems.These subsystems. are propositions concerning cross-levelformal Hypothesis3.2.2.2-2. In general,total entropy identities which can be demonstratedempiri- per unit cubic contentsincreases progressively cally. A few out of many such hypothesesthat along a distributorbetween the points of in- could be stated are listed below. These are put and output. selectedbecause theyare referredto in the next Hypothesis 3.2.5.2-1. If process A applied article. Many of the hypotheseswere suggested to any form of matter-energyalways precedes by the workof others,though usually the others processB (as, for example, convertingprecedes thought of them as related to one level only producing), variations imposed on the rate of and not as general systemshypotheses. It must process B by variations in the rate of process be rememberedthat when theyare applied to a A can be decreased by storing a supply (or specificsystem, allowance must be made for the "bufferinventory") of the outputs fromprocess disanalogies among systems.The variables in- A between the componentswhich carryout the volved show regular changes with level and two processes. type of system,and fromone individual system Hypothesis3.3-1. Up to a maximum higher of the same level and type to another. The than yet obtained in any living systembut less hypotheses are numbered by the same pro- than 100 per cent, the larger the percentageof cedure, distinguishingthe various subsystems all matter-energyinput that it consumes in in- processingmatter-energy and informationand formation-processingcontrolling its various sys- separatingstructure and process,which are em- tem processes, as opposed to matter-energy ployed in numberingthe sections of the next processing, the more likely the system is to article. survive. Hypothesis 1-1. In general, the more com- Hypothesis3.3.1.2-1. The frequencyof theout- ponents a systemhas, the more echelons it has. put signal of an input transducerincreases as a Hypothesis1-2. In general, the more struc- power function of the intensityof its input, turally differenttypes of members or compo- the form of the power function being iJ= nents a systemhas, the more segregation of K(S-So)n, where 4, is the frequency of the functionsthere is. output signal, 0 is the physical magnitude of Hypothesis 2-3. The more isolated a system the input energies,po is a constant,the physical is the more totipotentialit must be. magnitude of the minimum detectable or Hypothesis 3.1.2.2-1. When the boundary threshold input energies, k depends on the (except those portions containingthe openings choice of measurementunits, and the exponent for the ingestoror the extruder)of one living n varies with differentmodalities of the trans- system,A, is crossedby another,smaller, living ducer. or nonliving system,B, of significantsize, i.e., Hypothesis 3.3.3.2-1. In all channels C=W no smaller than the subsystemsor subcompo- log2 (1+P/N). That is, the maximum capacity nents of A, more work must be expended than (in bits per second) of a channel is equal to its when B is transmittedover the same distance band width times the logarithmof (1 plus the in space immediately inside or outside the ratio of the power of the signal to the power boundary of A. of the white Gaussian noise in the channel). Hypothesis3.2-1. An optimal mean tempera- Hypothesis3.3.3.2-2. There is always a con-

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 83 stant systematicdistortion between input and if R is less than C, there is a code which can output of informationin a channel. make the transmissionalmost freeof errors,and Hypothesis 3.3.3.2-3. In a channel there is as the systemmatures and associates, gaining alwaysa progressivedegradation of information practice, it gradually approaches such trans- and decrease in negativeentropy or increase in mission. noise or entropy. The output informationper Hypothesis3.3.4.2-5. If a transmitterwith an unit time is alwaysless than it was at the input. information transmission rate (in bits per Hypothesis 3.3.3.2-4. A system never com- second) of R is transmittingover a noisy chan- pletely compensatesfor the distortionin infor- nel with a capacity (in bits per second) of C, mation flowin its channels. and if R is greaterthan C, there is no way to Hypothesis3.3.4.2-1. As a systemmatures it encode the message so that the equivocation uses increasingly efficientcodes, e.g., codes (i.e., the uncertaintyas to what signalsthe trans- which require fewerbinary digits or equivalent mitterput into the channel which the receiver signals per input signal. These codes approach still has after he has received a message from but never actually reach the theoreticalmini- the channel) is less than R - C, but there is a mum number of symbolsrequired to transmit way to encode it so that the equivocation is the information. Efficientcodes also have the R - C + e, wheree is a positivefraction, less than followingcharacteristics: one and usually small, of C. Moreover, as a (a) Simple symbols are used for the most systemmatures and gains practice,it encodes in frequentmessages and more complex ones for ways so as to decrease the size of e. the less frequentones. Hypothesis3.3.4.2-6. As the noise in a chan- (b) The symbols are selected to minimize nel increases,a systemencodes with increasing confusionamong them. redundancy in order to reduce error in the (c) The symbolsare chunked in long rather transmission. than shortblocks. Hypothesis3.3.4.2-7. If messagesare so coded (d) Limitations on the transmitterof the that they are transmittedtwice, errors can be signal are taken into account. For example, if it detected by comparing every part of the first transmitshighly redundant signals,each one is message with every part of the second, but not coded, but some of the redundancy is re- which of the two alternative transmissionsis moved. correct cannot be determined. If they are (e) Limitationson the receiverare taken into transmittedthree times, they can be both de- account. For example, distinctionsto which the tected and corrected,by accepting the alterna- receivercannot react are neglected. tive on which two of the three transmissions Hypothesis3.3.4.2-2. If a transmitterof in- agree. formationis putting out informationcoded to Hypothesis3.3.4.2-9. As the amount of in- have H bits per symboland a channel has a formationin an input decreases (i.e., as it be- capacity (in bits per second) of C, then the comes more ambiguous), the input will more channel cannot transmitat a rate faster than and more tend to be interpreted(or decoded) C/H symbolsper second, though it is possible as required to reduce strainswithin the system. to encode the message so as to transmitat a Hypothesis 3.3.4.2-10. As the strengthof a rate of C/H-e symbols per second, where e strain increases,information inputs will more is a positive fraction,less than one and usually and more be interpreted(or decoded) as re- small,of C. quired to reduce the strain. Hypothesis 3.3.4.2-3. The quantity e (see Hypothesis 3.3.5.2-1. When a new informa- Hypothesis3.3.4.2-2) decreases as a systemma- tion input, B, is associated, usually more than tures and associates,gaining practice in coding once, with a familiar one, A, that elicits a information. certain output, B sooner or later becomes Hypothesis3.3.4.2-4. If a transmitterwith an capable of eliciting the same output as A. information transmission rate (in bits per Hypothesis 3.3.5.2-2. A systemassociates a second) of R is transmittingover a noisy given strain within it with motor acts which channel and all living channels are noisy- relieve it, so that such a strain comes to elicit with a capacity (in bits per second) of C, and the motoracts.

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 84 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48

Hypothesis3.3.5.2-3. A systemdoes not form purpose or goal; and (d) issuing a command associationswithout (a) feedback as to whether signal to carryout the action or actions. the new output relieves strainsor solves prob- Hypothesis 3.3.7.2-13. Decisions overtly re- lems, and (b) reinforcement,i.e., strain reduc- flectingvalues of a systemare made only at tion by the output. the highestechelon. Hypothesis3.3.5.2-4. Associationsestablished Hypothesis 3.3.7.2-14. A systemwhich sur- earlyin the life of a systemare more permanent vives generally decides to employ the least than those established later. costly adjustment to a threat or a strain pro- Hypothesis 3.3.5.2-5. In associating there is duced by a stress firstand increasinglymore an optimal ratio of correcttrials to errortrials, costlyones later. depending on the probability that specific Hypothesis3.3.7.2-16. The deciders of a sys- signals will regularly coincide in the system's tem's subsystemsand componentssatisfice (i.e., environment. In most experimental environ- make a sufficientlygood approximation to mentsand all natural environmentsthis proba- accomplishmentin order to survive in its par- bility is less than 1, and if association were to ticular environment) shorter-termgoals than occur with too few errortrials, a systemcould does the decider of the total system. not properly allow for probable future varia- Hypothesis 3.3.7.2-17. A systemcannot sur- tions in the appearance of signals. Since the vive unless it makes decisions that maintain probabilityof the signals regularlycoinciding the functionsof all its subsystemsat a suffi- also is nearlyalways greater than 0, if associa- ciently high efficiencyand their costs at a tion occurs with too many error trials, the sufficientlylow level that there are more than systemcannot profit soon enough from past enough resources to keep it operating satis- inputs. factorily. Hypothesis3.3.6.2-1. The longer information Hypothesis3.3.7.2-18. Systemswhich survive is stored in memory,the harder it is to recall make decisions enabling them to performat and the less likely it is to be correct,but the an optimum efficiencyfor maximum physical rate of loss is not regular over time. power output, which is always less than maxi- Hypothesis 3.3.6.2-2. Informationstored in mum efficiency. the memoryof a living systemincreasingly over Hypothesis 5.1-1. As the informationinput time undergoes regular changes- e.g., omis- to a single channel of a living system,measured sions, errorsor additions of noise, and distor- in bits per second, increases,the information tions- resulting from processes of selection, output, measured similarly, increases almost reorganizationwith other stored information, identically at firstbut gradually falls behind interpretation,and entropic decay of organiza- as it reaches a certainoutput rate, the channel tion. capacity, which cannot be exceeded in the Hypothesis 3.3.6.2-3. The removal from a channel. The output then levels off at that systemof informationrepresenting experience rate and finally,as the informationinput rate stored in the memory (as distinguishedfrom continues to go up, the output decreases the information constituting its template, gradually toward zero as breakdown or the whose removal is often fundamentallydamag- confusionalstate occurs under overload. ing or lethal) predictablyalters stochastic mea- Hypothesis5.1-2. Channels in living systems sures of the system'ssubsequent behavior, and have adjustment processes which enable them the degree of these changes increases as the to maintain stable, within a range, the simi- amount removed is increased. larityof the informationoutput from them to Hypothesis3.3.7.2-1. Everyadaptive decision the informationinput to them. The magni- is made in four stages: (a) Establishing the tude of these adjustment processes rises as in- purpose or goal whose achievement is to be formationinput rates increase up to and some- advanced by the decision; (b) analyzing the what beyond the channel capacity. These ad- informationrelevant to the decision; (c) syn- justmentsenable the output rate to remain at thesizing a solution selecting the alternative or near channel capacity and then to decline action or actions most likely to lead to the gradually,rather than to fall precipitouslyto

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 85

zero immediately whenever the information Hypothesis 5.1-10. As average input inten- input rate exceeds the channel capacity. sity increases, use of the omission adjustment Hypothesis 5.1-3. Among the limited num- processdecreases. ber of adjustmentprocesses which channels in Hypothesis 5.1-11. As input priority in- living systems employ as information input creases, average output rate increases. (Input rates increase are: omission, error, queuing, priorityis the probabilitythat a given message filtering,abstracting, multiple channels,escape, will be preferentiallytransmitted before other and chunking. (Omission adjustmentprocess is messages.) failingto transmitcertain randomly distributed Hypothesis 5.1-12. As input priority in- signals. Error adjustmentprocess is incorrectly creases, average processingtime decreases. transmittingcertain randomly distributedsig- Hypothesis 5.1-13. As the size of the input nals. Queuing adjustment process is delaying ensemble increases,the average processingtime transmissionof a sequence of signals which is increases. temporarilystored. Filtering adjustmentproc- Hypothesis5.1-14. As the size of the input ess is giving priorityin processing to certain ensemble increases, the use of all adjustment classes of signals. Abstractingadjustment proc- processes increases. ess is processing informationwith less than Hypothesis5.1-15. As the size of the output complete detail. Multiple channels adjustment ensemble increases, the channel capacity in- process is simultaneouslytransmitting informaw creases. tion over two or more parallel channels. Escape Hypothesis5.1-16. As the size of the output process is simultaneouslytransmitting informa- ensemble increases, the total processing time tion input. Chunking adjustment process is increases. transmitting meaningful information in Hypothesis5.1-17. As the size of the output "chunks" of signals rather than symbol by ensemble increases, the processing time per symbol.) Each of these processes applies symboldecreases. to random and nonrandom informationinputs Hypothesis 5.1-18. As average information except chunking,which applies only to non- input rate increases, the costs measured in random inputs with repetitious patterning to energy(ergs per bit); utiles (e.g., cents per bit); a systemthat can associate (or learn). Each of time (seconds per bit); or duration of the state these processeshas a cost in some sort of de- of the system(seconds before the state changes) creased efficiencyof information-processing. remain more or less constant for a period of Hypothesis5.1-4. Higher level living systems time and then finallyincrease rapidly, near the in general have the emergentcharacteristics of point where the performancecurve begins to more kinds and more complex combinationsof decrease fromthe maximumbecause the system adjustment processes than living systems at is overloaded. lowerlevels. Hypothesis5.1-19. As the percentageof total Hypothesis5.1-5. As average informationin- resourcesto meet costs (as defined in Hypoth- put rate increases,variation in output intensity esis 5.1-18) runs out, average output rate de- increases. creases. Hypothesis 5.1-6. As average information Hypothesis5.1-20. As the percentageof total input rate increases,the average processingtime resourcesto meet costsruns out, average output increases. intensitydecreases. Hypothesis 5.1-7. As average information Hypothesis 5.1-21. As the percentageof re- input rate increases, variation in processing sources to meet costs runs out, the size of the time increases. ensembledecreases. Hypothesis5.1-8. As average informationin- Hypothesis 5.1-22. As the percentage of re- put rate increases, the percentage of internal sources to meet costs runs out, output range channel capacity used in nontask communica- decreases. tion increases. Hypothesis 5.1-23. As the percentage of re- Hypothesis 5.1-9. As average intensity of sources to meet costs runs out, average process- input increases,up to a point average process- ing time increases. ing time decreases. Hypothesis 5.1-24. As the percentage of re-

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 86 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48 sources to meet costs runs out, use of omission, Hypothesis5.2-20. The decider of a system error, queuing, filtering,abstracting, multiple must resolve conflictsamong other subsystems, channels, and escape adjustment processes in- which signal their demands for autonomy,and creases. the suprasystem,which signals commands for Hypothesis5.1-32. The queuing adjustment compliance. processis employedmore frequentlythe higher Hypothesis 5.2-24. Conflictsamong various the peaks of informationinput overlap until sortsof alternativesare resolvedby a systemin such time as the length of the queue is greater differentways: than the local, short-termmemory capacity of (1) Between two mutually exclusive positive the system,and then the use of this adjustment goals, i.e., goals which elicit approach behavior, falls offrapidly in a confusionalstate. resolution is difficultif they appear to be of Hypothesis 5.1-33. As a corollary of the equal value, but choice is usually made quickly above, the effectivenessof the queuing adjust- withoutmuch vacillation. ment process is positivelycorrelated with the (2) With goals that are positive and negative amount of local, short-termmemory capacity. at the same time, approach occurs until the Hypothesis 5.1-42. A minimum rate of in- systemis near, then avoidance or movement formationinput to a systemmust be maintained from the goal occurs, and the systemtends to for it to functionnormally. vacillate for a time fairlynear but not at the Hypothesis5.2-1. As stressincreases, it first goal. improves system output performance above (3) Between two mutuallyexclusive negative ordinarylevels and then worsens it. What is goals, the systemvacillates from one to the extreme stressfor one subsystemmay be only other but tends not to make a decision. moderate stressfor the total system. Hypothesis 5.4.1-1. The rate of increase in Hypothesis 5.2-2. The greater a threat or the number of componentsof a young system stressupon a system,the more componentsof rises exponentiallyuntil it reaches a maximum, it are involved in adjusting to it. When no but this growthrate may be altered by several furthercomponents with new adjustmentproc- factors. essesare available, the systemfunction collapses. Hypothesis5.4.1-2. Growingsystems develop of Hypothesis 5.2-4. The range of stabilityof in the directionof: (a) more differentiation of de- a system for a specific variable under lack subsystems;(b) more decentralization of strain is a monotonicallyincreasing function cision-making; (c) more interdependence of the amount of storage of the input, and subsystems; (d) more elaborate adjustment boundaries; under excess strain, it is a monotonicallyin- processes; (e) sharper subsystem creasingfunction of the rate of output. (f) increased differentialsensitivity to inputs; more elaborate and patternedoutputs. Hypothesis5.2-5. There is an inertia to the and (g) matter-energyand information-processingvaria- Hypothesis5.4.1-4. If the rate of information bles which a systemmaintains in steady state, input into a systemfalls below a specificlower so that change in their ranges of stabilityis limit, normal growth of the systemis impos- much less disruptiveof systemcontrols if it is sible. undertakengradually. Hypothesis 5.4.3-4. Decentralization of de- Hypothesis5.2-7. When a barrierstands be- cision-makingin general increases the speed tween a systemunder strain and a goal which and accuracy of decisions which reduce local can relieve that strain, the systemordinarily strains. uses the adjustmentprocesses of removingthe Hypothesis 5.4.3-5. As decentralizationin- barrier,circumventing it, or otherwisemaster- creases,echelons or componentsof the system's ing it. If these effortsfail, less adaptive adjust- decider increasinglymake decisionswithout the ments may be tried, including: (a) attacking benefit of relevant informationexisting else- the barrier by energic or informationaltrans- where in the system. missions; (b) displacing aggression to another Hypothesis 5.4.3-6. The more decentralized innocent but more vulnerable nearby system; a system'sdeciding is, the more likely is there (c) reverting to primitive, nonadaptive be- to be discordant informationin various eche- havior; (d) adopting rigid, nonadaptive be- lons or componentsof its decider. havior; and (e) escaping from the situation. Hypothesis 5.5-1. The fartheraway a com-

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 87 ponent is fromthe point of traumato a system, set of alternatives. Were there no freedom to the less pathological is its posttraumaticfunc- choose from a set of alternatives, the corre- tion, and particularlythe less does the trauma sponding element would be a static,passive cog disturb its role in the system's hierarchical rather than an active unit. We suggest the processes. following general characterizationof organiza- tion. Consider a set of elements, each asso- Hypothesis5.6-1. If a system'snegative feed- ciated with its own set of alternatives. We back discontinuesand is not restoredby that now define a complexion as a particular set systemor by anotheron which it becomes para- of alternatives. There are, of course, as many sitic or symbiotic,it decomposes into multiple complexions as there are ways of selecting a components and its suprasystemassumes con- representative from each set of alternatives. trol of them. The set of complexions then has an entropy which is merely the sum of the entropies of the individual sets of alternatives so long as 15. CONCLUSIONS the elements do not interact. Complexion en- tropyis a maximum for independent elements. This analysis of living systemsuses concepts Maximal entropy,i.e., zero coupling, will be said of thermodynamics,information theory, cyber- to constitutethe condition of zero organization." netics,and systemsengineering, as well as the Ashby (1962, p. 255-257) also deals with this. classical concepts appropriate to each level. He says,speaking of what "organization" means The purpose is to produce a description of as it is applied to systems,"The hard core of living structureand process in termsof input the concept is, in my opinion, that of 'con- and output,flows through systems, steady states, ditionality.' As soon as the relation between and feedbacks,which will clarifyand unifythe two entities A and B becomes conditional on or facts of life. The approach generates hypoth- C's value state then a necessarycomponent of 'organization' is present. Thus the theory eses relevant to single individuals, types, and of organization is partly co-extensivewith the levels of living systems,or relevant across in- theoryof functionsof more than one variable." dividuals, types,and levels. These hypotheses He goes on to ask when a system is not a can be confirmed,disconfirmed, or evaluated by systemor is not organized: "The converse of experimentsand other empirical evidence. 'conditional on,' is 'not conditional on,' so the converse of 'organization' must thereforebe, as NOTES the mathematical theory shows as clearly, the concept of 'reducibility.' (It is also called 'sep- (1) Christie,Luce, and Macy (1952) call the physi- arability.') This occurs, in mathematical forms, cal form which the communication takes the when what looks like a function of several "symbol design," and the informationitself the variables (perhaps very many) proves on closer "symbolcontents." examination to have parts whose actions are (2) Bertalanffy(1956, p. 3) suggests that systems not conditional on the values of the other can be defined much as I define them, as "sets parts. It occurs in mechanical forms,in hard- of elements standing in interaction." And he ware, when what looks like one machine proves says that this definition is not so vague and to be composed of two (or more) sub-machines, general as to be valueless. He believes these each of which is acting independently of the systemscan be specifiedby families of differen- others. tial equations. "The treatment of 'conditionality' (whether (3) Rothstein (1958, p. 34-36) deals with the con- by functionsof many variables, by correlation straints among units of organized systems in analysis, by uncertainty other terms of entropy and communication as in- analysis, or by formationprocessing: ways) makes us realize that the essential idea "What do we mean by an organization?First is that there is firsta product space-that of of all an organizationpresupposes the existence the possibilities-within which some sub-set of parts, which, considered in their totality, of points indicates the actualities. This way constitute the organization. The parts must of looking at 'conditionality' makes us realize interact. Were there no communication be- that it is related to that of 'communication;' tween them, there would be no organization, and it is, of course, quite plausible that we for we would merely have a collection of should define parts as being 'organized' when individual elements isolated from each other. 'communication' (in some generalized sense) Each element must be associated with its own occurs between them. (Again the natural con-

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 88 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48

verse is that of independence,which represents search for common properties among diverse non-communication.) kinds of complex systems. The ideas that go "Now 'communication' from A to B neces- by the name of cyberneticsconstitute, if not a sarily implies some constraint,some correlation theory,at least a point of view that has been between what happens at A and what at B. proving fruitfulover a wide range of applica- If, for a given event at A, all possible events tions. It has been useful to look at the be- may occur at B, then there is no communica- havior of adaptive systems in terms of the tion from A to B and no constraint over the concepts of feedback and homeostasis, and to possible (A, B) couples that can occur. Thus analyze adaptiveness in terms of the theory the presence of 'organization' between varia- of selective information. The ideas of feed- bles is equivalent to the existence of a con- back and information provide a frame of straint in product-space of the possibilities." reference for viewing a wide range of situa- (4) In Cervinka (1948) the author very precisely tions, just as do the ideas of evolution, of distinguishes, at the group level, between a relativism,of axiomatic method, and of opera- concrete system,which he calls a "socius," that tionalism." is a single person in a group together with He goes on to assert that "hierarchic systems all his relationships, and a "groupoid," an have some common properties that are inde- abstracted system,which is a pattern of attach- pendent of their specificcontent.... ments of a single kind of relation selected by an "By a hierarchicsystem, or hierarchy,I mean observer, which interrelates a set of people. a systemthat is composed of interrelatedsub- (5) This definitionis consistentwith the usage of systems, each of the latter being, in turn, Weiss (1958, p. 140). Murray (1959, p. 24) pre- hierarchic in structure until we reach some fers the word "configuration" for an instan- lowest level of elementarysubsystem. In most taneous spatial arrangement of subsystemsor systemsin nature, it is somewhat arbitraryas components of a system (or "entity," in his to where we leave off the partitioning, and terms)and "structure"for an enduring arrange- what subsystems we take as elementary. ment. He distinguishes these clearly from an Physics makes much use of the concept of "integration" of recurrent temporal relations 'elementary particle' although particles have a of component processes, a patterning of tem- disconcertingtendency not to remain elemen- poral variables. tary very long. Only a couple of generations ago, the atoms themselves were elementary (6) This concept is not a product of our times. It particles; today, to the nuclear physicist they developed long ago. For instance, in the mid- are complex systems. For certain purposes of dle of the Nineteenth Century,Virchow (1862) astronomy,whole stars, or even galaxies, can wrote that the scope of the life sciences must be regarded as elementary subsystems. In one include the cellular, tissue, organism, and kind of biological research, a cell may be social levels of living organization. In modern treated as an elementarysubsystem; in another, times the concept of hierarchical levels of a protein molecule; in still another, an amino systemsis, of course, basic to the thought of acid residue. Bertalanffy(1956, p. 7) and other general sys- "Just why a scientisthas a right to treat as tems theorists. Even some scientists not ex- elementarya subsystemthat is in fact exceed- plicitly of such persuasion, who have perhaps ingly complex is one of the questions we shall been skeptical in the past, recognize value take up. For the moment, we shall accept the in such an approach. For example, Simon fact that scientists do this all the time, and (1962, p. 467-468) writes: "A number of pro- that if they are careful scientiststhey usually posals have been advanced in recent years for get away with it." the development of 'general systems theory' Leake (1961, p. 2076) sees value in the con- which, abstracting from properties peculiar to cept of levels for contemporarytheory about physical,biological, or social systems,would be biological organization. He writes: applicable to all of them. We might well feel "Life begins with complex macromolecules that, while the goal is laudable, systemsof such such as genes and viruses,and here the princi- diverse kinds could hardly be expected to have ples of physics and chemistrydirectly apply. any nontrivialproperties in common. Metaphor Macromolecules may be organized and inte- and analogy can be helpful, or they can be grated with many other chemical materials to misleading. All depends on whether the simi- form cells, which at Virchow's time were larities the metaphor captures are significant thought to be the basic units of life. Cells, or superficial. however, may be organized into tissues or "It may not be entirely vain, however, to organs, with specific integrationsserving their

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 89

specific functions. These tissues and organs tion sciences, cognition, and large parts of may furtherbe integrated into organisms,con- biology itself might conceivably be treated as stituting individuals such as human beings. special cases." Human beings, and indeed many other orga- A textbook of psychologyhas been written nisms, are capable of further integration and which embodies a conceptualization of a organization into societies. These societies in hierarchyof living systemslike that I advance turn may be integrated with a more or less in the present work. (Coleman, 1960). limited ecological environment." A recent presidential address of the Associa- The view is also well stated by de Chardin tion of American Medical Colleges included a (1959, p. 43-44): passage emphasizing the desirabilityof synthe- "The existence of 'system' in the world is sizing the medical curriculum around the con- at once obvious to every observer of nature, cept of the relations among levels of living no matter whom. systems. Hubbard (1967, p. 1079) wrote: "For "The arrangementof the parts of the uni- the medical student . . . the significance of verse has always been a source of amazement descriptionsat the molecular and submolecular to men. But this disposition proves itselfmore level must be presented in the context of their and more astonishingas, every day, our science relationship to the more complex organizations is able to make a more precise and penetrating of these same living systemsat the level of the study of the facts.The fartherand more deeply organ, the individual, and the family group." we penetrate into matter, by means of in- And there is widespread scientificand pop- creasinglypowerful methods, the more we are ular implicit recognition of hierarchical levels confounded by the interdependence of its of living systems.As one instance out of many, parts. Each element of the cosmos is positively six banners in one of the halls of the United woven from all the others; from beneath itself Nations Palais des Nations in Geneva depict by the mysterious phenomenon of 'composi- six levels of social organization. They say: tion,' which makes it subsistent throughout Family, Village, Clan, Medieval State, Nation, the apex of an organized whole; and from and Federation. above through the influence of unities of a (7) Herbst (1957, p. 28) makes it clear that one higher order which incorporate and dominate should make the level of reference explicit. it from their own ends. He says that often, in writing on group re- "It is impossible to cut into this network,to search, for instance, an author will change his isolate a portion without it becoming frayed level of reference from the leader (organism) and unravelled at all its edges. to the group and back to a group member "All around us, as far as the eye can see, (organism) again without explicitly referring the universe holds together,and only one way to the change. This produces confusing con- of considering it is really possible, that is, to ceptual ambiguity. take it as a whole, in one piece." (8) Illustrative of the similarities between the ap- Kaplan (1957, p. 12) has applied the concept proach outlined here and current thinking of a hierarchyof systemsto international rela- about electronic systemdesign is the following tions: "The same variables will be used at statement by Goode (1960, p. 15) concerning different system levels. The international the need to identifythe level of reference: systemis the most inclusive systemtreated by "Confusion . . . arises from consideration of this book. National and supranational systems the level of design. System design may be are subsystems of the international system. done: They may, however, be treated separately as "1) At the set level: that is, a radar, an igni- systems,in which case inputs from the inter- tion system,a navigation set. Any of these may national systemwould function as parameters. be designed on a system engineering basis, This holds also for subsystemsof nation states given a need and the necessaryanalysis of re- and even for personalitysystems." quirements. The Panel on Basic Research and Graduate "2) At the set of sets level: thus an airplane, Education of the President's Science Advisory a telephone exchange, a missile system,each is Committee of the United States in 1960 ap- itself a set of sets and is subject to system peared also to recognize value in a general design. systemsapproach (Seaborg, 1960, p. 1810). They "3) At the set, of sets of sets level: thus an wrote: "we suggest that there is great promise over-all weapon system, a telephone system, in such an emergingsubject as a general study an air trafficsystem, represent such sets of of complex systems in action, within which sets of sets." such very large questions as the communica- In a similar analysis Malcolm (1963, p. 4-5)

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions 90 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 48

distinguishes eight hierarchical levels in a ponent, assembly,subassembly, unit, unit com- large weapon system: system,subsystem, com- ponent, and part.

LIST OF LITERATURE

ARDREY, R. 1966. The Territorial Imperative. HURLEY, W. V. 1963. A Mathematical Theory of Atheneum,N. Y. the Value of Information. Rept. 63-3, May ASHBY, W. R. 1940. Adaptivenessand equilibrium. 1963. Port of New York Authority,Engineering J. Ment. Sci., 86: 478483. Department, Research and Development Divi- . 1960. Design for a Brain. (2nd ed. rev.) John sion, N. Y. Wiley, N. Y. KAPLAN, M. A. 1957. Systemand Process in Inter- 1961. Cybernetics Today and Its Future national Politics. John Wiley, N.Y. Contributionto the Engineering-Sciences.Foun- LEAKE, C. D. 1961. The scientificstatus of phar- dation for Instrumentation,Engineering and macology. Science, 134: 2069-2079. Research, N. Y. LEVINSON, D. J. 1959. Role, personality,and social . 1962. Principles of the self-organizingsys- structure in the organizational setting. J. tem. In H. von Foersterand G. W. Zopf (eds.), Abnorm. Soc. Psychol., 58: 170-180. International Tracts in Computer Science and MALCOLM, D. G. 1963. Reliability maturityindex Technology and Their Application, Vol. 9, (RMI)-An extension of PERT into reliability Principles of Self-Organization, p. 255-278. management. J. Industr. Engng., 14: 3-12. Pergamon Press, N. Y. MEYER, L. B. 1957. Meaning in music and in- BERTALANFFY, L. v. 1956. General systemstheory. formationtheory. J. Aesthet.Art. Crit., 15: 412- Yearbook Soc. Gen. Sys. Res., 1: 1-10. 424. CERVINKA, V. 1948. A dimensional theory of MILLER, J. G. 1965a. Living systems: Basic con- groups. Sociometry,11: 100-107. cepts. Behav. Sci., 10: 193-237. CHRISTIE, L. S., R. D. LUCE, and J. MACY, JR. 1952. 1965b. Living systems: Cross-levelhypothe- Communicationand Learning in Task-Oriented ses. Behav. Sci., 10: 380-411. Groups. Tech. Rept. No. 231, May 13, 1952. -. 1965c. Living systems: Structure and proc- Research Laboratory of Electronics, Massachu- ess. Behav. Sci., 10: 337-379. setts Institute of Technology, Cambridge. 1971a. Living systems: The cell, and Living COHEN, J. E. 1962. Informationtheory and music. systems: The organ. Curr. Mod. Biol., 4: 78- Behav. Sci., 7: 137-163. 256. COLEMAN, J. C. 1960. Personality Dynamics and 1971b. Living systems: The group. Behav. Effective Behavior. Scott Foresman, Chicago. Sci., 16: 302-398. DE CHARDIN, P. T. 1959. The Phenomenon of Man. 1972. Living systems: The organization. Harper, N. Y. Behav. Sci., 17: 1-182. GOODE,H. H. 1960. Intracompany systemsman- MuRRAY, H. A. 1959. Preparations for the scaffold agement. IEEE Trans. Engng. Mgmt., EM-7, of a comprehensive system. In S. Koch (ed.), 15: 14-19. Psychology: A Study of a Science, Vol. 3, HALL, A. D., and R. E. FAGAN. 1956. Definition of Formulations of the Person and the Social Con- system. Yearbook Soc. Gen. Sys. Res., 1: 18-28. text, p. 7-54. McGraw-Hill, N. Y. HARLOW, H. F., and M. K. HARLOW. 1962. Social NEYMAN, J., and E. L. SCOTT. 1957. On a mathe- deprivation in monkeys. Sci. Am., 207(5): 137- matical theory of populations conceived as 146. conglomerations of clusters. Cold Springs HARTLEY, R. V. L. 1928. Transmission of infor- Harbor Symposium on Quantitative Biology, mation. Bell Sys. Tech. J., 7: 535-563. 22: 109-120. HENDERSON, L. J. 1958. The Fitness of the En- NEYMAN, J., E. L. Scorr, and C. D. SHANE. 1954- vironment: An Inquiry into the Biological Sig- 1955. Statistics of images of galaxies with nificance of the Properties of Matter. Beacon, particular referenceto clustering. Proceedings Boston. of the Third Berkeley Symposium on Mathe- HERBST, P. G. 1957. Situation dynamics and the matical Statistics and Probability, vol. 3, p. theory of behavior systems. Behav. Sci., 2: 75-111. 13-29. PARSONS, T., and E. A. SHIS (eds.). 1951. Toward HUBBARD, W. N., JR. 1967. Janus revisited. J. a General Theory of Action. Harvard Univer- Med. Educ., 42: 1071-1080. sityPress, Cambridge.

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions MARCH 1973] THE NATURE OF LIVING SYSTEMS 91

PARKINSON, C. N. 1957. Parkinson's Law. Hough- VIRCHOW, R. 1862. Atome und Individuen, Vier ton Mifflin,Boston. Reden ilber Leben und Kranksein. Berlin. ROSENBLUETH, A., N. WIENER, and J. BIGELOW. 1943. (Trans. by L. J. Rather as: Atoms and indi- Behavior, purpose, and teleology. Philos. Sci., viduals. In Disease, Life, and Man. Selected 10: 18-24. Essays by Rudolph Virchow,p. 120-141. Stan- ROTHSTEIN, J. 1958. Communication, Organiza- ford UniversityPress, California. 1958.) tion, and Science. Falcon's Wing Press, Indian VON NEUMANN, J. 1958. The Computer and the Colorado. Hills, Brain. Yale UniversityPress, New Haven. SEABORG, G. T. (Chairman), Panel on Basic Research WEBER, M. 1947. The Theory of Social and Eco- and Graduate Education of the President's nomic Organization. (Trans. by A. M. Hender- Science Advisory Committee. 1960. Scientific progress and the Federal government. Science, son and T. Parsons.) Oxford UniversityPress, 132: 1802-1815. N. Y. SHANNON, C. E. 1948. A mathematical theory of WEISS, P. A. 1958. In R. W. Gerard (ed.), Con- communication. Bell Sys. Tech. J., 27: 379- cepts of biology: A symposium. Behav. Sci., 3: 423 and 623-656. 92-215. SIMON, H. A. 1962. The architecture of com- WHORF, B. L. 1956. Language, Thought, and plexity. Am. Phil. Soc. Proc., 106: 467-482. Reality. Technology Press, Cambridge.

This content downloaded from 128.103.149.52 on Sun, 2 Nov 2014 16:37:11 PM All use subject to JSTOR Terms and Conditions