Defining Life, Explaining Emergence

Claus Emmeche

1997

(online at http://www.nbi.dk/~emmeche/cePubl/97e.defLife.v3f.html)

Center for the Philosophy of Nature and Science Studies
Niels Bohr Institute
Blegdamsvej 17
DK-2100 Copenhagen
Denmark


Content:

Bibliographical Note
Abstract
Explaining things - introductory remarks
General attitudes and the standard view
Requirements for a definition
Life as the natural selection of replicators
Life as an autopoietic system
Life as a semiotic phenomenon
Downward causation
Implicitly well-defined general objects
Emergence as explanatory strategy: the observer reappears
Concluding remarks
Acknowledgements
Notes
References


Bibliographical note: Earlier versions of this paper were presented at the Princeton History of Science Workshop on "Growing Explanations", Princeton University, February 15, 1997; and at the meeting in the International Society for the History, Philosophy and Social Studies of Biology (ISHPSSB) in Seattle, USA, July 16-21, 1997. Different parts were published in a modified form as 1) Emmeche (1997): "Autopoietic systems, replicators, and the search for a meaningful biologic definition of life", Ultimate Reality and Meaning 20 (4): 244-264 [the original title was: "Is the definition of life important?"], and 2) Emmeche (1998): "Defining life as a semiotic phenomenon", Cybernetics & Human Knowing 5 (1): 3-17. The present web version below contains the complete argument of both articles. A further thoroughly rewritten version, accessible also for non-specialists, was made in collaboration with Charbel Niño El-Hani, and translated by him into Portuguese as a contribution to a book (this version can be found at www.nbi.dk/~emmeche/coPubl/99.DefVida.CE.EH.html).

Abstract. The strong version of Artificial Life claim that emergent computational patterns may not simply simulate life but realize the very phenomenon. This is one of several reasons why a definition of life is of interest. In this paper, it is argued that the received view of definitions of life in biology and philosophy is misleading. Generality cannot in general be dispensed with. Though criteria for adequacy of definitions are highly context-dependent, definitions of life are of a special nature, belonging to what is here called ontodefinitions. Separate definitions of life fulfilling a set of relevant criteria exist and belong to distinct paradigms of theoretical biology. Emergence is implicit in these. The paper investigates if emergence entails `downward causation' (several kinds are defined), and asks if computational models can represent higher orders of emergence. Finally, a comment is made on the role of Wittgenstein's philosophy in understanding the nature of explanation and definition in science.

Explaining things - introductory remarks

Modern biology explains life, living beings, as a highly organized material entities composed of cells composed of molecules and as results of a long process of evolution by ripe with emergent structure. Today we hear the claims of new artificial life, that it is possible to mimick evolution not only by gene manipulation and other biotechniques, but by computer and robot techonology as means to create similar or even identical instances of living organization realized in other media. In what sense will these other patterns, processes, entities be alive? Will it effect the way we conceive and explains the living organisms we already know? Is it merely a `matter of definition' whether we call the artificial creatures alive or not -- and what would it really say to define life and to change one's understanding of biological phenomena? Let's start reflecting on the very notion of explanation.

By explaining things, we change them. They do not remain the same in our conceptions. Explanations should satisfy our quest for understanding. There are, of course, forms of understanding that do better without explanation. Jokes should not be explained unless somebody makes a request, and explaining the point and even why it was funny can be a demolishing undertaking. Being a good thing to pursue in science, called upon in the court-room, often required in the upbringing of children and in everyday life, we should, however, be careful about the use of explanations. In science, there is a certain flavour of reduction to the very notion of explanation that underlies the dualism between explanation of mechanisms in the natural sciences and understanding of meaning in the humanities.[1]

This paper is very far from addressing all questions pertaining to the semiotics of explanation in science; it focuses on one single aspect; the explanatory role of ontodefinitions: For the moment, these can be thought of as certain very broad categories -- such as matter, life, mind, or society -- which are not simply denoting huge phenomenologies in a very vague manner, but also refer to some categories of the so-called modern scientific and evolutionary world picture, whose symbolic and cognitive content science has contributed to deepen. The ontodefinitions in science, of which I shall only consider the concept of life, are very basic for the paradigmatic character of scientific activity. They belong to the metaphysical component of a paradigm of a disciplinary matrix in Kuhn's sense. They are almost mixed explanations and definitions. They define what scientists are looking for (thus constitute what is relevant and what entities experiments should deal with), and at the same time, the provide some basic understanding of the very nature of these objects, a narrative, an explanatory story of some kind. This is all implicit in a given paradigm. The ontodefinitions are not considered by the scientists to be important or to have anything to do with the usual everyday experimental activity. They don't care. We shall see that biologists are typically reluctant to define life in explicit terms; nonetheless, such definitions exists within distinct paradigms of contemporary biology. I will consider the role of these explanations in the scientific quest to universally understand life as a coherent emergent phenomenon, as pursued in traditional biology and within one of the `sciences of complexity': the interdisciplinary field of Artificial Life, which has brought some attention to the idea that life is an intrinsically semiotic phenomenon (even though instead of `semiotic', the term `informational' is more often used). This investigation is in part motivated by some general assumptions about scientific activity as explanatory.

Standard ideas in philosophy of science[2] inform us that different explanatory strategies are used in different sciences, that is, in the physical sciences, the explanations are causal or causal/mechanical (whatever that means) whereas the life sciences may use, in addition, functional or teleological explanations. Similarly, within the social sciences, functional explanations may play a role, but more important are intentional explanations, also known in cognitive science as the intentional stance (i.e., explaining the behaviour of an agent, whether a machine, a human being, or an animal, as-if the agent has particular intentions, purposes, etc.). Views may differ about the derived or non-fundamental status of some of these forms, and about whether only one form, typically the causal one, should be preferred also in biology and sociology, or whether, e.g., teleological explanations are fundamental for biology. However, for the work done in such interdisciplinary areas of complex systems research, Artificial Intelligence and Artificial Life, these `received' schemes do not give an adequate pictures of what is really going on in these fields and what kinds of explanatory approaches are at stake here.

Within Artificial Life [3], with its special research agenda of trying to study `life-as-it-could-be' (not simply life-as-we-know-it in its Earthly instances), the focus is put on the `natural' generation of complex objects, that by their very development or evolution (as represented by a computational model) crosses the boundaries between traditional classes of objects as constituted by physics, biology, and psychology; e.g., models for the origin of life from a purely chemical `soup' of elements; models for the generation of a multicellular organism from a single fertilized cell; or models for the creation of internal representations (boldly called `mental') of the outer world within an artificial agent such as a wall-following robot with sensors, motor, and a neural network. By these approaches, alife research crosses the boundary between the area in which only causal/mechanical explanations are used and areas where functional and intentional explanations are allowed or even requisite: the kingdoms of teleology and mentality. That such borders are crossed is at least a tempting interpretation of one of the new dimensions of philosophical importance of this research, and is consistent with its own emphasis of emergent phenomena. Typically, one can identify two levels of interpretation of the model,[4] the lower level's pre-specified rules for local interactions (the generalized genotype, or GTYPE), and the global level which exhibits a structural or behavioral PTYPE (generalized phenotype) that emerges as a result of non-linear interactions, and thus, one may even study the "important feedback mechanism between levels in such systems".[5]

The basic promise of Artificial Life research[6] is to use the new computational resources (or new robot techniques, or new biotechnology) to construct completely new and bona fide forms of life (software, hardware and wetware forms) in order to overcome what Carl Sagan has called the fundamental handicap of biologists, that they basically knows only one particular instance of life, namely life on Earth, which constitutes fundamentally `the same' form of life on the biochemical and metabolic level, governed by the same mechanisms of inheritance, the same types of metabolic patterns, etc. By constructing quite new instances of life, alife research promise add to our primitive Zoo of archetypes of lifeforms and thus be able to explain or understand what is universally true about life and what is simply contingent upon the particular way life evolved on Earth (e.g., we might extend the notion that all life require cells with DNA and RNA to a notion of life as component systems containing partial self-descriptions, which may or may not be based on an organic chemistry of carbon chain molecules).[7]

This intuition is twofold: (a) What we can construct we are also able to explain. (We can imagine an isomorphy between constructing and explaining, for instance, a detailed procedure for assembling a machine may give us enough information to construct an explanation of its workings in the form of an algorithmic description of the rules for its change of state). This is probably a very common sentiment in physical science.[8] (b) Complex things in nature construct themselves as wholes via long processes of local interactions between simple entities; this emergence of wholes (or collective behavior of units), should be mimicked in our explanations. Instead of `top-down' reductive explanations, complexity research provide `bottom-up' explanations of emergent phenomena. Even though these explanations are still reductive (in the methodological sense that one can in principle show exactly what is going on from step to step in a simulation of, e.g., the evolution of new species), the complexity of the system makes prediction impossible, i.e., computational short-cuts (to predict the future state) cannot be found for every case, and hence some simulations are computationally irreducible. (c) In addition to these two intuitions, a third notion of information-controlled development or information-based complexity is often invoked; namely that such self-adapting, self-reproducing and self-organizing systems at that level of complexity require that some part of the system regulate, control, or modulate the other parts.[9]

The intuitions behind the computational alife research programme show, first, that even though it is admitted that strict reduction as explanatory strategy for complex living things may not be possible, the phenomena are still considered to be open for explanation of another kind; second, these explanations have an important narrative aspect. In fact it can be argued, that any explanation in science has a narrative aspect in that the story to be told (the explanans) must have, as it were, the right logical and semiotical power of generating a believable representation of the story's subject matter (the explanandum). The explanation, comprising both explanans and explanandum, should in a narrative form (of which deductive logic is considered here as a highly specific kind) lead a listener to a particular kind of understanding of the phenomenon. Explanations are a certain kind of narrative rules of a science-specific game of language, that is, the game of generating a representation of a phenomenon by logical mechanisms such as deduction, induction, abduction, or by more elaborated computational and hermeneutic procedures such as construction and interpretation of computer models of emergent phenomena. Explanations are constructive: In explanations, we reproduce a system by producing another one.[10] Though this is true of any explanation, the extensive use of computer simulation and related visualization has made this aspect particular salient.

It is tempting to ask what is the relation between explaining life by constructing computational models of lifelike behavior and defining the so created `emergent' patterns as true instances of life? Is Artificial Life redefining the notion of living systems in biology, or does it for the first time give a universally valid definition of life? In the year 2000, the term biology will celebrate its second centennial, and the modern idea of a unified science of living systems is at about the same age. Biology before Artificial Life was (and still is) considered as an autonomous science with its own standard methods, theories, and basic assumptions about its subject matter -- living systems, life, organisms. Should we really believe in the explanatory strategy of the `strong thesis' of Artificial Life, that life in a genuine sense (not just representations, but the very phenomenon) can be artificially created in vitro, or in silico, so that AL research contributes to explore life from a much more universal point of view? How can one be sure that life simply can be defined as "the emergence (in any kind of medium) of complex structures with certain life-like properties"? What makes this notion counterintuitive to some biologists? Do organisms have to be material? And why have biologists been so reluctant to give clear definitions of life that could be used as a measure to hold up against the simulation when the alifer claims it to be a `real' living thing? We may even ask: What may be the meaning of the fact that all attempts to formulate a satisfactory definition of life have failed?

As I hope to show, the very presupposition that it has failed, seemingly accepted by a majority of biologists, is misleading, based on wrong premises. Biologists have for long taken the Aristotelian notion of a definition for given, in which something is defined (`by genus and species') as a member of a class (here: living beings) if and only if it shares with the other members a permanent set of essential (necessary and sufficient) properties that can be listed and checked (e.g., organization, nutrition, growth, development, reproduction, irritability, susceptibility to illness and death). Such a conception of a definition of life (the standard view) leads to problems, and these problems (vagueness, special exceptions, etc.) have led some biologists to think that in biology, we cannot generally define life. This is in conflict with some fairly good intuitions (such as Sagan's) that all life (on Earth) do share some fundamental properties. This intuition is historically and empirically well grounded. I shall argue that the problem is not just to define life -- as this has already been done implicitly by 20th century biology in two important ways. In order to evaluate the alife claim one must grasp the nature of the definition and its cognitive role, and its relation to the notion of emergence, a term often used quite informally. The strategy here is to show that the standard view of the definition of life is flawed; that two effective implicit definitions of life are in use in contemporary biology (and a further one is possible); and that these definitions constitute life as an emergent phenomenon, but only implicitly.

General attitudes and the standard view

The view that defining life is a futile endeavour is very common among biologists, even though this complete and unreflective refusal of the very question does not constitute what we below shall call the standard view. Though possibly interesting, it is hardly necessary to do a systematic survey with questionnaires to practising researchers within biochemistry, molecular, cellular and evolutionary biology, in order to observe that most scientists are extremely sceptical toward attempts to make clear definitions of living beings - their objects of study. They simply assert (with some justification) that a definition is of no use in solving the various experimental puzzles of normal research. Go into a DNA lab or a molecular biology department, attend a research seminar, and ask about what definition of life the researcher takes as his or her point of departure, and you will be met with an indulgent smile.

On the one hand, this sceptical refusal of the very question may surprise the outsider of science, given the common idea that science must afford clear and logically consistent definitions of all concepts employed. "Why then", the outsider may ask, "should this not apply to such a concept denoting the very subject matter of the entire field of research?". However, one should recognize that experiments are not done on "living systems" in abstracto, but on Escherichia coli, Caenorhabditis elegans, Drosophila, mouse, and other of the biologists' pet creatures that serve as the preferred experimental model systems. These `systems' are, of course, considered as living, but normally it is neither a problem to see this, nor does it pose a problem for the concrete experimental research really to "define" such systems. This is the pragmatic answer as to why biologists usually do not consider the definition question as an important one: Who cares, so long as we can distinguish dead from living fruit flies? - and of course, we can, and even if, in a particular setting, we cannot, definitions will not help.[11]

On the other hand, from the point of view of biology as a general field of inquiry, it is surprising to realize that so few attempts had been made to reflect systematically and critically upon the nature of living systems, as one should expect it to be a general concern of biology to contribute to a clear and universal understanding of the make-up of life's processes and its organisation. This `bias' towards experimental biology may simply reflect economical priorities due to expectations of higher technological spin-offs and other benefits from research, but compared with the field of physics, long known for its tradition of theoretical research and striving for universal validity of its theories, one can indeed be puzzled about why so few researchers in biology are doing theory and endeavouring at a comprehensive understanding of life.[12]

Definitions of life are seldom discussed in depth or even mentioned in biology textbooks[13] or dictionaries,[14] a reflection of the general sceptical and empiricist attitude toward what is taken as "merely theoretical" or metaphysical speculation in contrast with the "facts" of experimental research. A remark made by Ernst Mayr is representative for this attitude: "Attempts have been made again and again to define `life'. These endeavours are rather futile since it is now quite clear that there is no special substance, object, or force that can be identified with life."[15] This reflects the typical stance among biologists towards definitions of life.

However, Mayr, known as one of the main contributors to `the modern synthesis' in neodarwinian evolutionary theory, is by no means an empiricist, and he transcends the pure refusal to define life. Thus from the citation just given he immediately proceeds by acknowledging a kind of definition of life: "The process of living, however, can be defined. There is no doubt that living organisms possess certain attributes that are not in the same manner found in inanimate objects. Different authors have stressed different characteristics, but I have been unable to find in the literature an adequate listing of such features."[16] Mayr then informs the reader that his own list is probably both incomplete and somewhat redundant, but it will "illustrate the kinds of characteristics by which living organisms differ from inanimate matter". The list, which I will not discuss in depth here, has as its key words: 1. Complexity and Organization; 2. Chemical Uniqueness (referring to the high specificity of biochemical macromolecules); 3. Quality (whereby Mayr makes a contrast between physics as a quantitative science and the biological world as a world of qualities, individual differences, communications systems, interactions in ecosystems, etc.); 4. Uniqueness and Variability; 5. Possession of a Genetic Program (with the notable remark that "Nothing comparable to it exists in the inanimate world, except for manmade computers."[17]); 6. Historical Nature (of the taxonomic categories, e.g., species, which cannot be considered as classes in the logical sense); 7. Natural Selection; 8. Indeterminacy (under this tag Mayr includes unpredictability because of randomness, stochastic perturbations, complexity and "emergence of new and unpredictable qualities at hierarchical levels"). Therefore, one may conclude that on the one hand, Mayr thinks that definitions have to be essentialistic ones and that attempts to define life by one single "essence" or crucial characteristic are futile, but on the other hand, it is indeed possible to define life as a process by a very qualitative and possibly redundant list of eight properties (or nine, if we include emergence and irreducibility[18]).

Thus, the central assumptions in Mayr's approach are the following set of claims which we may call the standard view of the definition of life (SVDL) because it seems to constitute `the received view', not only in modern evolutionary and theoretical biology but in most branches of biology:[19]

(i) Life as such cannot be defined, thus a clear definition is missing.

(ii) The question of defining life is not important for biology.

(iii) However, living processes may be defined, or at least approximately demarcated from inorganic processes through a list of characteristic properties (non-essentialism).

(iv) Difficulties in delimiting such a set of properties are recognized, but are not considered to be serious (given the pragmatic nature of the definition and the sceptical view of the use of any definition at all). Particular living beings may not hold all properties given, so the list may not be a list of necessary and sufficient properties; it may be more vague or redundant.[20]

(v) Even though life is a physical phenomenon, biology deals with systems of such a vast complexity that we cannot in practice[21] hope to reduce it to physics. Should we list the crucial properties of living processes, then complexity, organization, and genuine biological ones such as self-reproduction and metabolism (cf. Mayr's list) would be inevitable.

The central claims are (i) and (ii), and most biologists will usually not feel committed to make further elaborations of the consequences of this stance. This may be caused by a misconceived view of definitions as being by necessity Aristotelian, or `essentialistic' in character. Biologists of today are emphatically non-essentialist and non-vitalist. They shy general claims of the nature of life that have to them the slightest stain of vitalistic ideas or connotations. This attitude may be decent if one faces attempts to substitute quasi-scientific, pseudo-holistic, or intuitive notions of life for the normal scientific view of life as highly complex and specific physical systems open for biological, chemical and physical inquiry. However, it is only a sad consequence of the defeat of vitalism that some biologists still conceive the foundation of modern biology to be mechanistic, i.e., rooted in classical or quantum physics, rather than having its own autonomy as a science with a foundation, or a set of paradigmatic ideas, that are better described as being organicistic.[22] Organicism in twentieth century biology has been a sort of philosophical counter-movement, opposed to vitalism and mechanism.[23] Today it has a curious status in the American and Anglo-Saxon tradition of philosophy of biology as a neglected position or one with only a historical interest; even though we can find and reconstruct it as `a spontaneous philosophy' of most biologists (and a more reflected such one in the writings of, e.g., E. Mayr, R. C. Lewontin, S. J. Gould and other contemporary evolutionary biologists).

One can also say that the attitude of subscribers to the SVDL is this: "Don't talk too much of definitions, enough has been said." The neodarwinian philosopher Michel Ruse provides a clear example of the SVDL: "Life. This, the distinguishing feature of organisms, is best thought of as involving some kind of complex organization, giving an ability to use energy sources for self-maintenance and reproduction. Efforts to find some distinctive substance characterizing life have proven as futile as they have been heroic. The one thing which is clear is that any analysis of life must accept and appreciate that there will be many borderline instances, like viruses. Inconvenient as this may be for the lexicographer, this is precisely what evolutionary theory would lead us to expect."[24]

Though the SVDL is `standard' for most biologists of various breeds; theoretical biologists attracted to Artificial Life, as well as computer scientists, physicists, and mathematicians within this field, share a more open attitude to the possibility of (re)defining biological life. In a sense, the very idea of studying life-as-it-could-be enforces an interest in more general definitions of life. Within the neodarwinian tradition, we find John Maynard Smith (also interested in alife[25]) as a exemption to the unreflected view that it is not interesting to define life. He surpasses the SVDL in his attempt to give a general definition. Maynard Smith (1986, pp. 1-8) argues for two criteria, namely (1) metabolism ("although the forms of living organisms remain constant, the atoms and molecules of which they are composed are constantly changing"), and (2) functions ("the parts of organisms have `functions', that is, the parts contribute to the survival of the whole"[26]). Maynard Smith goes on to discuss their interdependence and relation to biological evolution. Even though metabolism is a thermodynamic process of maintaining a `dissipative structure' through intake of low-entropy energy and dissipation of high-entropy energy (like a vortex in a river or a burning flame), the maintained structure of even the simplest organism is enormously more complex, and the flow of energy is controlled, a phenomenon that has a lot to do with the second feature and with evolution by natural selection. Maynard Smith does not simply give a list -- he relates the listed features internally within the theoretical framework of evolutionary biology. He seems to be in harmony with Mayr's view of biology as consisting of two kinds of research, one dealing with the `proximal' causes of e.g., metabolism, as in biochemistry and physiology, and another dealing with the `ultimate' causes or functional adaptations of organisms -- genetics and evolution theory. Both must be seen as integrated within the modern version of darwinism so that life can be defined, according to Maynard Smith, by the possession of those properties that are needed to ensure evolution by natural selection, i.e., "entities with the properties of multiplication, variation, and heredity are alive, and entities lacking one or more of those properties are not".[27] As we shall see, this definition is close to being identical with the first of two candidates for a theoretically satisfying (though implicit) definition of life in theoretical biology.

Requirements for a definition

We have noticed that many biologists are sceptical about the use of definitions of life and do not appreciate previous attempts which they think have failed; and that several biologists do not think it is possible in principle to define life in a precise, mathematical or exact way. Definitions are usually considered as speculative and of no use in guiding practical experimental studies.[28] These attitudes are never followed by clear statements of what would be the requirements of a possible set of acceptable valid definitions of life. We could ask: What requirements should a definition of life meet? Are such requirements specific to biology or general for any scientific term? Should a definition of life reveal the deepest nature of (biological) life?

These questions can be answered in two steps. The first is from a point of view of biological science -- an interpretive interest in understanding the most general objects of biology, life itself. To list the requirement is an attempt to make explicit the paradigmatic nature of any specific ontodefinition of life, in order to advance our understanding of organisms as material, informational, and semiotic phenomena. A second step is to use this explicitation to remove some false presuppositions about the role of definitions in science.

The requirements of a definition of life[29] can be given as the following demands that can be justified on theoretical and pragmatic reasons: generality; coherence and non-vitalism; comprehensive elegance; and specificity:

(a) A valid definition of life should be general so as to encompass all possible forms of life, not just the contingent products of darwinistic evolution on this planet. Life may not on other planets have its genetic material stored in DNA molecules or it may not have a metabolism based on proteins with enzymatic function, but it will probably have both a metabolism of some kind and genetic memory of some kind. Even though we do not know for sure, it is very hard to imagine (at least for 20th century biologists) forms of life which do not have (or, which are not parasitic on other forms with) a kind of genotype-phenotype duality, where the genotype is the genetic code or memory, and the phenotype is the manifest organism belonging to some general type. Generality must be ontological as well as epistemic, that is, the definition must in principle cover all possible forms of life in the universe, and it must not be a simple reflection of the particular epistemic or disciplinary framework in which the definition is given (cf. Sagan, 1973, who has drawn attention to the tendency for each biological discipline to have a particular definition of life).

(b) A definition of life should not involve notions which stands in opposition to what we already know of living things and their inorganic components, i.e., it should be coherent with the general understanding of living systems based on biological research, and also with modern physics and chemistry, and based on this tradition it should be non-vitalistic in having no reference to occult powers of life, some supernatural directing forces or whatever, even though it does not have to entail an ontological reductionism.

(c) A definition of life should have what we might call a conceptual organizing elegance, i.e., it can organise a large part of the field of knowledge within biology and crystallize our experience with living systems into a clear structure, a kind of schematic representation that summarizes and gives further structure to the field. The role of a definition of life is not simply the role of definitions of more or less technical terms within specific sub-fields (e.g., what defines "the primary immune response" of the immune system, or "the peptidyl-tRNA binding site" of the ribosome); rather, its role is to give the general object of study within the broader area of biology a clear profile, to organize our cognitive models and theories of living systems in a unifying and coherent way (as a kind of rational root metaphor),[30] and to distinguish the scientific study of life from other sorts of inquiry, such as philosophical investigations of the existential and phenomenological aspects of human existence in a society, or the scientific study of cognition and the human psyche, or the study of physical matter. The definition of life should be so elegant that it can hold true for multicellular organisms with an immune system, such as the vertebrates, as well as for single cells, and one should be able to relate specific components of life at the subcellular level, such as ribosomes, to its general properties. It should give a comprehensive view of life compared with matter, mind, and society, and it should enable us to comprehend the internal unity in the biologic diversity of life.

(d) Though able to give an idea about any kind of system that does have the ability to live, metabolize, replicate its own kind (or what else might be considered relevant properties of life) a definition should be specific enough to distinguish life from obviously non-living systems (this follows from (c)). Of course, a moment's reflection may indicate a circularity in the demand for a definition to demarcate life from non-life by appeal to what is taken as `obviously' living or non-living. We can have different intuitions, and our evidential knowledge may fail. How can we be sure that a crystal, that can grow and eventually multiplicate, may not, to a certain extent, be alive? We could manufacture various definitions that meet every person's different intuitions -- pantheistic, materialistic, dualistic, or whatever -- about the nature of life. Are we just building conceptual schemes by imposing on the malleable world of experience? Such a view would be too pessimistic and would in effect block the way of further inquiry. The world is not infinitely malleable, and we should not dismiss the genuine and detailed knowledge of living systems[31] we have won during this century. A more pragmatic view takes for granted that there are real things (and real differences between specific real categories of things), whose characters are entirely independent of our opinions about them, and that we can come to know these realities through the methods of science. Our understanding of living systems is based on fallible, but nevertheless scientific knowledge of the distinctive characteristics of living cells and organisms.

Thus we can require that a definition of life is general enough to deal with life as a universal phenomenon, not just Earthly carbon-life, and not just life-as-a-thermodynamic-system or life-as-genetic-processes, etc.; that it is coherent with current knowledge of biology and physics; that it makes no appeal to vitalistic forces; that it shows some conceptual elegance and a cognitive organizing capacity; and that it is specific enough to catch the basic primary characteristics of biological life, no more, no less. If we can find or explicate, within the paradigm (or some sub-paradigms) in modern biology, a notion of life that fulfil these criteria (an ontodefinition), a little part of the constructive metaphysics of science has been unveiled.

It has already been noted that such a set of requirements are not demanded for definitions of every scientific term; most terms represent much more specific kinds of objects or processes. For instance, in biological classification, we wish to define every species, genus, order, class, etc., and we demand more confined criteria that specify an organism's potential class-membership of a biological taxon (e.g., a species) based on a set of `essential' (necessary and sufficient) characteristics. In evolutionary studies of speciation, we can define a species as `an individual' in which the relation between the species and its members is a part-whole relation (not an element-class relation as in classification).[32] In molecular biology, we may (preliminarily and operationally) define a specific polynucleotide involved in protein synthesis as the behaviour of an isolated `factor' whose presence changes the translational properties of the ribosome. However, when defining life, we wish to demarcate a very broad class of processes, a very general and organised mode of physical systems (different from culture, society, mind or matter). The cognitive scientific function of this definition is of another kind, so it would be a fallacy to demand the same level of operational or conceptual concreteness of such a definition.

A definition of life should not by necessity reveal the final reality of `life'. Science is an open form of inquiry different from other forms, and here, we are dealing only with a specific kind of scientific definitions of life looked upon as real biological systems. From a scientific point of view, there may be other kinds of inquiry that reveal other aspects of life that are not explained by natural science, not even approached by biological research, except perhaps in a very indirect way (aspects of what we in ordinary language call life, such as life as a normative concept of representing something good, or something to be preferred rather than the dead state of things, the feeling of appreciation for living nature, etc.). These aspects may motivate research, they are important to us in a lot of ways; from an existential point of view, they may be even be the most important ones. Whether the individual scientist subscribes to an instrumentalist, a realist or a pragmaticist philosophy of science, he or she will usually recognise the importance of a clear and comprehensive understanding of the distinctive features of biological systems, even if one does not believe that science by necessity reveals the final reality of life.

One has to be careful taking the second step, i.e. to reform our view of the role of definition in science. On the one hand, we should emphasize that a lot of technical and theoretical terms in science still have to be defined according to what Lakoff and Johnson criticize as "the objectivist account of definition", whereby a category is defined (in terms of set theory) as a set of inherent essential properties of the entities in the category.[33] On the other hand, we need a broader view of the specific cognitive role of various kinds of definitions in science encompassing how scientists interact with and comprehend the objects of their field, including objects of the most general kind. As science is still a specific activity distinct from (but not privileged with respect to) other activities of ordinary life, we will not propose a view of scientific definitions as being merely prototypical or metaphorical (in the sense of Lakoff and Johnson), though prototypes and metaphors are indispensable as cognitive tools in scientific reasoning. The sharp demarcation between "the context of discovery" (with its vague, intuitive and metaphorical kinds of reasoning and understanding) and "the context of justification" (where all basic terms are extensionally defined and all arguments meets the standard requirements of logic) has for long been a target of critique in the philosophy of science, so time should be ripe for considering the consequences of this critique for accounts of the kinds of definition and categorization in science. Even though science and metaphysics are distinct enterprises, there is interdependence and mutual exchange, as revealed in debates concerning definitions of the most general kinds of objects of the special sciences, e.g., definitions of matter, life, psyche, consciousness, culture, sociality, etc. We have invented the word ontodefinitions as a term for the integrative but often vague and implicit character of these notions within a given scientific paradigm -- and for the theoretical attempts to define more explicitly these very broad ontological categories. When dealing primarily with science (and scientist's grasp of the most general kinds of objects), we can simply observe that `defining a term' is not a question of one single methodological procedure or one specific game of language, and in matters where science and ontology meet, we should cross between the Schylla of rigorism and the Charybdis of murky vapourings.[34] Now let's take a closer look on the specific definitions of life.

Life as the natural selection of replicators

A significant aspect of the first definition is its peculiar status within biology. The idea, that life basically can be defined as the natural selection of replicators or self-copying entities is often ignored or not recognized as a definition at all, due to the dominating SVDL and its sceptical pragmatic attitude. Even though neglected this definition somehow lives an implicit life of its own within evolutionary biology. It is easy to make explicit, and once this is done, it would likely be accepted by the majority of evolutionary biologists, who are used to think of life not on the level of the individual organism, but as lineages of organisms connected by the processes of reproduction and selection.

This definition can be formulated as a generalization of statements about the kind of entities that undergo variational change, i.e., evolve by natural selection.[35] Maynard Smith's 1986 contribution (see above) can be generalized this way. According to this definition, life is a property of populations of entities that: (1) self-reproduce, (2) inherit characteristics of their predecessors by a process of informational transfer of heritable characteristics (implying a genotype-phenotype distinction[36]), (3) vary due to random mutations (in the genotype), and (4) have the propensity to leave offspring determined by how well the combination of properties (inherited as genotype and manifested as phenotype) meets the challenge of the environmental selective regime.

This formulation can be made even more abstract if we emphasize that by "genotype" and "phenotype" we do not necessarily refer to particular genes made of DNA or organisms made of cells, but to any kind of "replicators" and "interactors". The term replicator comes originally from the zoologist Richard Dawkins, 1976, who thought evolution primarily to proceed by selection at the level of genes (replicators), which by the very process of replication preserved their structure in time. For Dawkins, life on Earth began with the appearance in the primordial soup of molecules that could replicate, i.e., catalyse the production of their own kind by template copying processes. This led to the evolution of cells and multicellular organisms as a "survival machines" for their replicating genes, i.e., the perpetual information sequences written in the nucleotides of DNA molecules. Thus Dawkins' defining characteristic of life amounts to the natural selection of more and more effective replicators.[37]

However, while Dawkins embedded his view of gene selection in a reductionist metaphysics of which only the replicating structures were the real entities in evolution, while organisms were like ephemeral, transient epiphenomena, the philosopher David L. Hull wanted to broaden the ontological framework of the theory of evolution, by introducing the general terms replicators (any entities that pass on their structure directly by replication), interactors (any entities which produce differential replication by means of directly interacting as cohesive wholes with their environments) and lineages for the entities at work in darwinian evolution: "A process is a selection process because of the interplay between replication and interaction. The structure of replicators is differentially perpetuated because of the relative success of the interactors of which the replicators are part. In order to perform the functions they do, both replicators and interactors must be discrete individuals which come into existence and cease to exist. In this process they produce lineages which change indefinitely through time. Because lineages are integrated on the basis of decent, they are spatiotemporally localized and not classes of the sort that can function in laws of nature". (Hull, 1981, p.41).

It is sad that Hull's contribution (apart from philosophical discussion of the species concept) has been much overlooked in discussions of definitions of life, because it seems to fulfil pretty well definitional requirements (a) to (d) we discussed above. Though we obviously cannot know for sure, it is highly conceivable that all life in the universe evolves by a kind of Darwinian selection of interactors, whose properties are in part specified by an informational storage that can be replicated. No non-physical forces are involved in this process, yet the very notion of natural selection and replication (the transfer of sequence information that specifies biological activity of macromolecules) seems to be specific for biological entities;[38] lineages of interactor-replicator entities are not described in the textbooks of physics. This definition is simple, elegant, general, and crystallizes our ideas of the general mechanisms of the creation of living systems within an evolutionary perspective.

Richard Dawkins has been most successful as popularizer of his idea that life is a kind of naturally selected informational systems of replicators. A problem with Dawkins' idea (in contrast to Hull's) is its bias toward a purely informational conception of life. We normally consider life to be both form and matter -- something with both informational-organizational and material-physical aspects -- but in Dawkins' version the definition puts much emphasis on the informational aspect, viz. the replicators as self-propagating patterns of information. This may mislead weak souls to a nearly Platonic concept of life as simply defined by any instantiation of some specific set of abstract informational properties (such as an entity's ability to replicate its own informational description).

Such instantiations can be found in Artificial Life where the computational and man-mediated representations of life-processes are seen not just as simulations, but realizations.[39] This "strong artificial life" view implies that we can synthesize genuine examples of life, simply by having strong enough computational support and good enough (informational) criteria to decide whether our simulations are just approximations or whether they realize enough properties to be considered alive. The strong thesis depends upon (1) a notion of life as a very general process that can be instantiated -- for example, as systems of replicators -- in various material media (which may be correct if we allow for the specific dependence of each of these forms of life on their specific medium), and (2) a claim (similar to functionalism in the philosophy of mind) of medium-independence of life, i.e., the idea that the realized life is essentially `the same' in all media -- characterised only by its informational aspects independent of the material `substratum' supporting these processes (which seems to be incorrect[40]).

An indication of the connection between defining life as a set of replicating entities undergoing natural selection and the tendency to claim the possibility of having genuine life realized simply as digital informational structures in a computer (that may include representations of mutations, replication and natural selection) is the very term "replication" which is often taken to be identical to "self-reproduction". This is not the case. RNA and DNA molecules replicate, either in the test tube or in living cells, but cells grow and divide which is the true complicated process of self-reproduction (which includes molecular replication). One can both simulate replication and self-reproduction processes on a computer. However, one cannot realize true self-reproduction in a computer, because in contrast to self-reproduction of cells in nature, which is complete (in the sense that all information needed for the process is contained in the self-reproducing structure itself), "simulated self-reproduction" is not complete in that sense because it depends upon the information contained not only in the simulated replicating entities, but also on the information in the embedding program and computer that runs it (as pointed out by Kampis, 1991; Kampis and Csányi, 1987). There is a lot of structure in the cell that constitutes the very material causal conditions for having -- as part of the cell (or interactor) -- a `replicator'. This structure is self-maintaining, or a self-producing component system and thus points to the next definition of life as autopoietic systems.

How does this first ontodefinition of life relate to the notion of life as an emergent phenomenon? It does not in itself explicate the notion of emergence in a standard sense (in which "a property of a complex system is said to be `emergent' just in case, although it arises out of the properties and relations characterizing simpler constituents, it is neither predictable from, nor reducible to, these lower-level characteristics"[41]). However, we noted that Mayr included emergence (combined with the notions of indeterminacy and unpredictability) in his list of features of "processes" characteristic for life. Defining life as the natural selection of replicators seems to imply a notion of life as an emergent phenomenon in two respects.

First, it presupposes semiosis in the form of genetic systems with `digital' codes for information transfer, i.e., systems characterized by properties that supervene on but cannot be reduced to physical properties (the genetic code -- as well as other semiotic systems -- is a good case of a system of properties that are supervenient on its molecular or chemical supervenient base).[42] We can simply formulate this as follows: (i) Life as replicators demands the existence of semiosis (perhaps just in very simple forms). (ii) A semiotical system has properties (such as stable coding relations) that are emergent with respect to their physical base properties. (It is hardly possible to imagine any semiotical system in which there is just one level of description and no emergent properties). (iii) Thus, life as replicators could, as an ontodefinition, be reformulated within a more comprehensive biosemiotical framework.

Second, during adaptive darwinian evolution, genuine new properties appear that cannot (even in principle) be predicted in advance (due to random mutations) or explained merely by physical or chemical theories.[43] Thus by defining life as the natural selection of replicators, it is implicitly supposed to be an emergent phenomenon.

Life as an autopoietic system

We have seen that the informational (cryptosemiotical) definition of life as natural selection of replicators helps to explain the origin of the functional characteristics of organisms as interactors. Now remember that besides function, Maynard Smith mentioned metabolism as a basic feature of life. The closed network of metabolic components within a cell is a point of departure for understanding the second candidate for a definition of life: Life as an autopoietic system, as defined by Humberto R. Maturana and Franscisco J. Varela. Autopoiesis literally means self-production or self-creation, and is a term for the "self-defining", "circular" organization (organizationally closed but structurally, i.e., materially and energetically, open) of a living system (such as a cell), consisting of a network of component metabolites that produces the very network and its own components plus the boundary of this network.

The definition of life as autopoietic is distinct from the first definition in several respects: (1) It has been deliberately invented as part of a general, abstract theory of life, a theory that attempts not only to catch the biological phenomenon of life in its most general sense as a contribution to a theoretical biology, but also to give a biologically founded epistemology -- the distinctions the observer makes (among living systems and any other units) are reflected in the theory from its very beginning. This epistemological feature of the theory is interesting, but transcends the scope of this paper.[44] (2) It rejects the notion of genetic or biologic information as something intrinsic to the autopoietic system; rather, information is seen as being ascribed to the system from an observer's point of view. Any form of intrinsic teleology (or, by implication, semiosis) is also rejected. (3) Another characteristic feature of the theory and definition of life is that (even if referential relations in the strict intentional sense is prohibited within the theoretical framework) it illuminates a self-referential aspect of life, i.e., the closed topology of internal relations, the idea that living systems can "only be characterized with reference to themselves"[45] -- which is reflected in the self-referential character of the definition of autopoiesis and the whole theory:

The definition of life as autopoietic systems (Maturana and Varela, 1980, e.g., pp. 78-84, 97, 135) is formulated in a mechanistic framework. The living cell is an autopoietic machine, which is "a machine that is organized (defined as a unity) as a network of processes of production, transformation and destruction of components that produces the components which: (i) through their interactions and transformations regenerate and realize the network of processes (relations) that produced them; and (ii) constitute it (the machine) as a concrete unity in the space in which they (the components) exist by specifying the topological domain of its realization as such a network". Furthermore, "the biological phenomenology is the phenomenology of autopoietic systems in the physical space and a phenomenon is a biological phenomenon only to the extent that it depends in one way or another on the autopoiesis of one or more autopoietic unities."[46] Autopoiesis is an all-or-non property; a system cannot be `more or less' autopoietic.[47] Man-made machines are not autopoietic in that they do not by themselves generate their constituents. According to the theory (and its many, precise, interrelated, and unusual definitions of terms), such central biological phenomena as evolution, self-reproduction, replication, are phenomenologically secondary to the constitution of autopoietic units in the physical space.

Maturana and Varela, 1980, reject the mere list-approach to the definition of life as theoretically unsatisfactory. Furthermore, the theory of autopoiesis will not (in contrast to biochemistry and molecular biology) be a theory of the properties of components of living systems (ibid., p. 113). It is a theory of the concatenation of processes of production that constitute autopoietic systems, not the application of physical and chemical notions to biological phenomena. It is a theory of biological organization that explains the autonomy of living systems, not a chemical theory of the `structure' of the cell's component macromolecules. Finally, it is a strictly non-representational, non-semiotical and non-teleological theory of life, any notions of a `genetic programme' that somehow `codes for' or `represents' information of the whole organism, of signs or signals that are being `interpreted' by the organism, or of `purposes', `functions', or `goals' of subsystems of a living being, are considered to be purely metaphorical, and not, according to the theory, how such systems actually operate in themselves.

Interestingly, this definition has to a large extent been neglected by biologists as well as philosophers of biology, but for other reasons. Its emphasis on life's non-teleological nature may seem counterintuitive to many biologists. For example, molecular biology focuses on `molecular recognition' and structural-functional descriptions of genes and proteins, with a quasi-teleological way of thinking of genetic material as possessing a `blueprint' for the development of an adult organism from a fertilized egg. A major reason for neglect of the autopoietic definition of life, however, is probably the intricacy of arguments in the theory, its perplexing style, as well as its metaphysical dimension, which is sometimes considered as having solipsistic inclinations, though such an interpretation can hardly be justified from a closer reading.[48]

The autopoietic definition of life is clearly not just about a particular Earthly instance of DNA-based life. We could think of autopoiesis in virtual, nonphysical spaces. On the other hand, it coheres with current biological knowledge and makes no reference to spiritual powers. Despite its painstaking style, the theory offers a particular and logically consistent way of seeing life and thus has a remarkable cognitive organizing capacity. Though one can question (Fleischaker, 1988) whether other instances of autopoietic systems than biological ones exist, the definition of autopoiesis seems to be specific enough to capture a very fundamental feature of biological life -- its autonomy and closed organization.

Is an autopoietic system by implication an emergent one in relation to its physical constituents? Obviously, the theory of autopoiesis adds something to the physicalist intuition of causal closure in the physical domain: organizational closure in the biological domain. One may portray the theory as a anti-teleological strategy to account for the teleological nature of living systems by explicating in which sense such systems have emergent properties not reducible to properties of physical systems. In fact, the phenomenological framework in which the theory was formulated emphasizes exactly that though the origin of autopoietic organization may be explicable "with purely mechanistic notions (...) once the autopoietic organization is established it determines an independent phenomenological subdomain of the mechanistic phenomenology, the domain of the biological phenomena." Thus, "one phenomenological domain can generate unities that define a different phenomenological domain, but such a domain is specified by the properties of the new different unities, not the phenomenology that generates them." (Maturana & Varela, 1980, p.116). Again, though the concept of emergence is not explicated, the general idea is implicitly present in the theoretical context of this definition of life.

Both definitions of life fulfil the criteria (generality; coherence; comprehensiveness; and specificity) given above for this kind of ontodefinition of terms for the most general kinds of scientific objects that are at the same time the objects of ontology. The two definitions belong to two separate paradigms of biology. "Life as the natural selection of replicators" is rooted in neodarwinian evolutionary biology, which is nowadays perceived as being based on a thoroughly molecular genetic description of heredity (in spite of the metaphorical complex of informational terms such as `the genetic code', `biological information', etc.[49]). "Life as autopoietic systems" belongs to a separate minor (but important) branch of theoretical biology with origins in systems theory, cybernetics, neurobiology and a positivistic view of scientific explanation. Both definitions implicitly construe life as an intrinsically emergent phenomenon.

Life as a semiotic phenomenon

Despite eventual incommensurability between the theoretical terms and different metaphysical commitments of the first two ontodefinitions, partial communication between their respective advocates is indeed possible. It is natural to ask if it is possible to integrate them into one even more general one. In fact, important work has already been done in this direction.[50] This work will probably in the future be integrated with biosemiotics as a new and promising paradigm of theoretical biology (or philosophy of biology), allowing us another way to perceive life, not as based on the organisation of molecules, but as based on the communication of signs in nature.[51] In biosemiotics, the focus of attention it is not the natural selection of replicators or the operational closure of an autopoietic system, but the sign-links and interpretants of various semiotic agents on all biological scales, from molecular recognition to cellular self/non-self-distinction, from the molecular semantics of gene expression and regulation to the semantics of inter-organism communication from butterflies to elephants, from individual cognition to the swarm intelligence of ants and humans. We may even expect a major recasting of our view of life within the next decade from this sign-theoretical perspective. However, before we can glimpse this development, some fundamental issues have to be resolved and the first set of questions is concerned with the nature of the biosemiotic `paradigm' itself: Could biosemiotics really be a new paradigm for theoretical biology that -- just like the modern synthetic theory of evolution -- can guide experimental research in specific sub-disciplines and provide a coherent framework for the study of life? Or will it rather be a meta-theoretical reflection about the conditions of possibility for doing biological research (e.g., that the nature of biological knowledge is dependent upon the semiotic character of the object of study)? Or will it be a kind of new philosophy of nature, in which the world is seen from the very beginning as a meaningful living universe with inborn potential for creative generation of new signification? These questions cannot yet be answered.[52]

The second issue concerns implications of viewing life as a semiotic phenomenon. If we define life as biosemiotic processes, these have to be specified in terms of organisms (i.e., sign-interpreters) and their functions. Lets us say that life is defined as functional interpretation of signs in self-organized material code-systems making their own umwelts. This definition seems to imply that information (signs, or meaning) is conceptually primary; while organisms, metabolism and evolutionary replication are secondary with respect to the semiotical processes. When we apply the semiotical concepts to natural systems, this is often taken for granted. However, in order to bridge the gab between (physical) nature and (semiotical) culture[53], we have to develop a theory of the causal nature of sign-interpretation that can account for the generation of the so-called original meaning (not just ascribed observer-dependent meaning) as part of the natural activity of physical systems under specific boundary conditions. A similar search for such a theory is going on in cognitive science and the neurosciences.[54] As we have not yet seen convincing accounts of the emergence of sign-functions in a purely physical system (cf. Emmeche 1994b), such a theory will probably be closely dependent on a future biological understanding of the origin of life (living cells, metabolism and the semiotical machinery of the genetic memory), as it is precisely here we can detect the first primitive semiotic systems. As biological life (as implied by both of the first two ontodefinitions) is functional -- i.e., dependent upon system-maintaining relations between parts which are not living (such as a single molecule of sugar or a protein) and wholes which are living (such an amoeba or a cat) -- it remains to be seen precisely how the causal nature of the part-whole relations is constituted.

If the cell (e.g., the first cell in the primordial soup) is an emergent entity, or has emergent properties (such as metabolism, self-reproduction), this entails a kind of `upward causation' from the physical collection of individual macromolecules (lacking such properties) to a functional whole. One has to answer if this process entails just the emergence of functionality, as many biologists would be inclined to say, or if functionality is an inherently semiotical phenomenon, so that for any kind of complex system (above the von Neumann threshold) this kind of part-whole relation can only be realized by being based upon a sign-interpretation capacity of the very system itself. In case of the cell, its metabolism is based on proteins, and these are partly specified informationally by the genome (the primary sequence of amino acids), partly being self-assembled in the folding process following the protein synthesis. Here, functionality, as revealed by molecular biology, is not `purely physical'; it is biological, and in that respect, also semiotical: In the genome of an eukaryotic cell there is sequence information specifying a set u of roughly 50,000 - 100,000 specific species of protein molecules of which maybe 10,000 may be in use in a given cell type. The cell's DNA-based selective construction of one specific protein (from the set u) for synthesis -- out of the astronomically huge number of possible proteins of a given size -- is really the original meaning of `biological specificity' as the term for the distinctive character of biochemical reactions, which was later identified as being based upon the `biological information' in the genome. The secret of biological complexity lies in that `semiotically correctly' specified selection. The biosemiotical claim is that by casting molecular biology as just `biology with chemical and physical methods', we tend to forget the intrinsically biological meaningfulness in the cell as the basic semiotical unit. The cell as a semiotical whole causes (by a kind of `downward causation') another much more ordered distribution of materials and kinds of molecules than what would have existed without this emergent semiotical unit.

Downward causation

This downward causation should not be interpreted in a certain strong sense; it does not necessarily mean that the `causal closure' of the physical universe is disturbed `from without' (as if we are invoking a semiotic analogy to a vitalistic `life force') or that the physical laws are being violated or something of that sort.[55] However, to talk about the causal closure of the physical universe is somehow to take a transcendental "God's eye view" on the whole of nature. We cannot attach much pragmatic scientific meaning to such a proposition as `the whole nature is causally closed'.[56] As finite semiotical beings we should not invoke such demons of traditional physics as a complete micro-determinism. A more modest stance will emphasize the necessity of choosing a particular frame of description and particular observables in the quest for understanding life, and here we are forced to surpass the purely micro-deterministic stance. Downward causation seems to be a problem only for an ontology that allows only strictly efficient causation, because efficient downward causation surely leads to contradictions.[57] But semiotical causation is different. It involves other causal modes (or modes of explanation), almost forgotten but still helpful to us, from the Aristotelian tradition, namely material, formal and `final' [functional] causation.

It is true that emergence of new entities on a higher level imply the existence of downward causation (compare Küppers, 1992). However, one can distinguish three versions of downward causation, each with distinct ontological assumptions; here briefly characterized as follows:

(1) In strong downward causation, an entity or process at a higher level may causally inflict changes or effects on entities or processes on a lower level, and the higher level entity is considered to have a substantial difference from lower level entities. The organizational aspect is a necessary but not sufficient condition of the higher level entity: By its emergence, an ontological change in substance takes place. Thus, the higher level is held to constitute its own substance; it does not merely consist of its lower level constituents (this could be called constitutive irreductionism). Vitalism in biology and dualism in philosophy of mind may invoke strong downward causation of some version.

(2) For medium downward causation it is not allowed higher level phenomena to influence directly on lower level laws. The higher level entity, such as a cell or a psyche, is a real substantial phenomenon in its own right, and this entity acts as constraining conditions (a kind of formal cause) for the emergent activity of lower levels. The higher level states, already realized, are constraining conditions for the coming states.[58]

(3) In weak downward causation the higher level is seen as an organizational level (not a substance), characterized by the pattern, the structure or form into which the constituents are arranged. The higher level entity, for instance a biological cell, consists of entities belonging to the lower level (constitutive reductionism). This is not physical reductionism; the forms of the higher level are believed to be non-reducible (form realism). It does not interpret boundary conditions as constraining conditions; rather, the higher level form can (in terms of the theory of dynamical systems in physics) be seen as a stable or chaotic attractor in a phase space where the individual states (points in the state space) of the system is given by the configuration of the system's lower level entity properties and the dynamical equations that rule the time evolution of the system.

In computational artificial life models, we typically have a case of weak emergence (weak downward causation) of perceived `patterns' of movement (of, e.g., cellular automata), which from the very start was in a sense pre-specified by the state transition rules of the automaton. It is more contentious if we can have stronger forms of downward causation. A problem with weak downward causation as an possible candidate for describing the causality in emergent phenomena such as life and mind is that we can only in a very metaphorical way talk about `the phase space' of, e.g., biological species of psychological thoughts. Of course, we cannot specify such states physically or apply the phase space description in any literal sense here. For genuine emergent phenomena, our intuition is rather that new rules of `dynamics' are being invented on higher levels of description (e.g., with the origin of human language, the linguistic rules of grammar was invented; with the origin of the first living organisms, the `rules' of the genetic code was invented and the `sequence space' of DNA-bases and amino acid strings was invented on a much higher level than the `physical state description of particles' in dynamical systems theory).[59] In the computational model, it seems as if no real newness can be created; all is fixed from the start of the simulation by the model's pre-specified rules and the initial conditions, and micro-determinism (or `syntactical determinism') reigns. An attractor is a great idea analytically to come from one level (the micro) to the next, but in order to go further, one must either investigate a whole dynamics of attractors (which is mathematically very complicated), or allow introduction of new observables `from without' and thereby simplify the system description.[60] The guess is that we are either faced with a possible fundamental limit of computational models (that they cannot generate second and higher order forms of emergence and stronger forms of downward causation), or that the very notion of causal interaction between entities at different levels (even if we allow for a more broad set of causal relations than the efficient midro-deterministic one) is incoherent and needs revision.

Implicitly well-defined general objects.

The discussion of the two more established definitions suffices to support a central point. Many philosophers and biologists have tended to believe that it is a simple fact that all attempts to formulate a satisfactory definition of life have failed. Some biologists would even add that this is completely insignificant for research, and one should simply study concrete cells and organisms and give particular molecular, functional and evolutionary explanations of these systems; any attempt to define life would result in overly general concepts. I have argued that this opinion can be contended for several reasons.

First, even though one embraces the standard view of the definition of life, one does not necessarily deny that `living processes' may be defined, demarcated or characterised in a general way, for example, through a list of shared properties of living beings, even though such a list may be vague, incomplete, redundant and may not constitute a set of universally necessary and sufficient conditions.

Second, definitions of scientific terms cannot be restrained to one single type of definition (e.g., operational, mathematical, ostensive, Aristotelian, ontological or whatever) because of the manifold character of research. Talking about biological life is talking about a very general set of objects -- in fact the whole subject matter of the biosciences -- and we should not be too rigid in our demands for precise definitions, especially where the cognitive and theoretical function of a definition serves as a root-metaphor for the whole field (which may motivate further inquiry to make it more specific and general). This is the case for what we tentatively have termed ontidefinitions. In contrast to the standard view (including the idea of a of a "fuzzy" borderline between living and non-living processes), we have seen that life can nevertheless be defined with rather good theoretical precision.

Third, biology in the 20th century has not only been empiricist and "fact"-oriented, but has given us rich conceptual tools to construct a coherent picture of at least some of the universal properties of living systems (cells, multicellular organisms, and systems of such organisms), such conceived within an evolutionary frame as evolved, highly organized, adaptive systems with some autonomy and specific informational properties, i.e., with properties that are emergent, but no less material than chemical and physical properties. In this sense, organisms are genuine ontological units, and well-defined as objects of biology.[61]

Fourth, even though it is controversial how to interpret the results of Artificial Life research gained so far -- with respect to (a) clarifying how life indeed could-be, i.e., distinguish the contingently shared common properties of life on Earth from the universal generic properties of life of any form, and (b) more philosophically, deciding if some simulations may be considered as realizations -- this field represent a set of inspiring approaches and methods not only to synthesize vehicles, animats, new kinds of self-replicating molecules, and new virtual universes of complex informational forms, but also to win a more general understanding of the principles of complexity that is alive, and perhaps a general feeling for the laws of form that universally constrain the biological process of living.[62]

When we investigate the possibility of defining life as a semiotic phenomenon, as a system of signs mediated by interpreting organisms, we should remember that a definition of life is itself a sign -- a sign of the quest for simplicity, comprehension and scientific understanding. Wittgenstein (1953, SS 432) remarked: "Every sign by itself seems dead. What gives its life?--In use it is alive. Is life breathed into it there?--Or is the use its life?". Only as abstracted and isolated from the practice of biology will definitions of life fail to give us insight. As dead metaphors they have been handed over to philosophical analysis. However, as we have seen, good definitions of life do exist and enjoy a life of their own. Definitions should be used, and a definition of life can be used as a condensed paradigmatic expression for a whole view of living beings.

In this sense, the three ontodefinitions of life are very general implicit explanations of what kind of physical systems the living ones are. Whereas definitions have often been considered as explications of particular concepts required for an explanation, with these definitions there is built-in, so to speak, a certain understanding or explanation of life. In this sense they are paradigmatic: they provide the biologist with a way to `see' and life and explain particular instances.

It may appear to some as a contradictio in adjecto to speak of living phenomena as specified in the three ontodefinitions as `implicitly well-defined general objects' of biology. Let me stress that by well-defined I do not suggest that for instance the problem of borderline cases will not appear (as cases of conceptual vagueness may reflect the existence of vague boundaries in Nature[63]); they are well-defined only relative to the criteria for adequacy given above. This does not mean that such objects cannot be more clearly defined if we untangle some of their implicit properties. One such was emergence.

Emergence as explanatory strategy: the observer reappears

As mentioned in the introduction, the notion of emergence has after a long period of oblivion[64] been revitalized at the end of this century by the sciences of complexity, focusing on the complex emerging properties of life and mind. Considering something to be emergent is no longer perceived as something mysterious, in conflict with a scientific world view[65], implying metaphysical dualism, etc. We saw that in Artificial Life, one of the basic intuitions was that we can computationally imitate emergent processes of construction observed in Nature as the creation of new wholes on higher levels of organization. Thus, instead of top-down reductive explanation of constituent structure, one can as a complement search for bottom-up explanations of emergents. What we can construct, we should be able to explain, because in contrast to alchemy, the constructions seen in Nature are completely material, and the corresponding computational constructions must be, on the basic level, completely algorithmic. This is the idea.[66]

What form does such an explanation have? It is not `hypothetical-deductive' or predictive; it is often acknowledged that the complexity of (the model of) the system makes `prediction' impossible. This is due to (a) the computationally irreducible nature of the mathematical model (computational short-cuts cannot generally be given, so one has to go through the whole sequence of states, which is hardly an act of prediction), and (b) the fact that prediction involves comparison between a real measured system and a model, and in order to computationally answer questions about future states a model with deterministic chaos requires definite (arbitrarily high precision) initial conditions, which cannot be provided by measurement of parameters of the real system due to the normal condition of measurement uncertainty.[67] Further, in Artificial Life focus is on systems with nonlinear interactions, because in contrast to linear systems (obeying the superposition principle that the simpler parts can be analyzed independently) nonlinear systems must be treated as wholes in which the behaviour of the whole is `more than' (or better: different from) the behavior of the parts; cf. Langton, 1989, p.41: "Behaviors themselves can constitute the fundamental parts of nonlinear systems -- virtual parts , which depend on nonlinear interactions between physical parts for their very existence.(...) It is the virtual parts of living systems that Artificial Life is after: the fundamental atoms and molecules of behavior".

The alife form of `bottom-up' explanation may more appropriately be called `interpretational emergent explanations', or `jumping to the conclusion', because two levels of interpretation is involved in the model, and the observer interpreting the model is crucial for establishing the emergent phenomenon. This is often ignored: The as-if-emergence of higher-level patterns of behavior in the model (often represented visually on the computer screen as a 2-D virtual world) based on low-level computation of interactional primitives (the rules representing local interactions) are often seen as simply realizing emergent behavior, where one tends to forget that these patterns are not real in any trivial sense.[68] If there is nothing intrinsic biological to the emergent phenomena of the model, the emergence may simply be in the eye of the beholder.[69]

In order to access more precisely whether something is emergent, a more formal framework (based on the theory of categories[70]) has been proposed by Nils A. Baas. It is interesting in this context, because the function of an observer in establishing an emerging property is here explicitly recognized as a requirement at any level. Baas considers his idea as a step towards a general theory of hierarchies, complexity, emergence and evolution. These four interrelated phenomena (the `hyperstructure' of the theory) are always found in biological systems, but also in the computational ones of Artificial Life and in dynamical systems. Whenever we encounter life, it must be hierarchically organized; hierarchies are the things that have had the time to evolve from simple to complex structures; and complexity is here used in the algorithmic sense of needing a long `programme' for the specification of the system or a long route of computational development. Life cannot do with just one macro-level/micro-level distinction; hierarchies are what make complexity manageable through several levels of organization. Evolution by natural selection is the process which gives rise to new levels. By evolution the environment, as it were, acts as an observer that `sees' or `acts upon' higher level properties, thereby establishing recurrent forms of interactions within and between the different levels.

According to Baas, 1994, for something new to be created we need some dynamics or better interaction between the entities. But to register that something new has come into existence, we need mechanisms to observe the entities. So emergent properties must be observable, but they appear because of the system of interactions among the lower level objects, not because of observation. Baas does not specify the nature of the observing subject or the `observational mechanism' because only general and formal requirements for emergence are intended. The process of emergence of properties on several levels may be considered as a result of a series of abstract construction processes, similar to mathematical constructions. Given a set S1 of first order structures, one can, by some kind of observational mechanism Obs1(S1) obtain or `measure' the properties of the structures at this level. The S1's can then be subjected to a family of interactions, Int, using the properties registered under observation (this could be a dynamic physical process). Hence one gets a new kind of structure, S2 = R(S1, Obs1(S1), Int), where R stands for the result of the construction process. The interactions may be caused by the structures themselves or imposed by external factors. Obs is related to the creation of new categories in the systems. S2 is a second order structure, a new unity whose properties may now be observed by another observational mechanism Obs2, which may also observe the first order structures it consists of.

Baas defines P as an emergent property of S2 if and only if P belongs to the set Obs2(S2), and P does not belong to the set Obs2(S1) -- which may be interpreted as saying that the whole is more than the sum of the parts.

The idea of emergence as a function of interaction and observation is indicated in the figure. Baas distinguish between different types of emergence, (a) deducible/computable emergence, which means that there is a deductional or computational process D such that P can be determined by D and Obs1(S1); and (b) observational emergence, which is the more profound type, characterized by the condition that if P is an emergent property, it cannot be deduced as in (a). Type (a) emergence clearly indicate, that the defining characteristic of an emergent property P -- that it belongs to Obs2(S2) but not Obs2(S1) -- does not entail that it could not be determined by Obs1(S1) in an explanation using D. This is close to the idea of Kincaid, 1988, that the irreducibility of a higher level theory does not entail that lower level theories, with respect to some questions, cannot explain higher level phenomena.

Baas exemplifies deducible emergence by non-linear dynamical systems where simple systems interact to produce new and complex behaviour: phase transitions, broken symmetries and many types of engineering constructions where coupled components interact in known ways so that we can calculate the new compositional properties of the system. Chaotic dynamical systems is considered as a borderline case of (a) and (b). A genuine example of observational emergence in a formal system is according to Baas found in Gödel's theorem, i.e., in the fact that in some formal systems there are statements which are true, though this cannot be deduced. Here observation is the truth function (ascribing the truth function value "true" to a string of symbols which is not deducible from the axioms and rules of inference of the formal system itself). Starting from first order calculus -- which is complete in the sense that every true statement can be deduced from the axioms -- one can add further axioms to cover the theory of arithmetic, and this `adding' is a kind of `added interactions' among well-formed expressions which creates emergent properties. Though Gödel's theorem as observational emergence may at first appear as a rather negative result for constructive purposes, it is not, because Baas have "incorporated the observational mechanisms in such a way that they can usefully be taken into account in further constructions -- even if they are not deducible" (ibid.). Also the property of membership of the Mandelbrot set and most Julia sets may be observationally emergent. Furthermore, the semantic non-compositionality of a language (for instance natural language) would imply that the meaning of sentences in such a language were observationally emergent.

Thus, Baas concludes that even in formal, abstract systems (including models of life-like processes) profound kinds of emergence may occur. In Baas' general theory, both the traditional micro-deterministic upward causation and the more controversial downward causation (that may even occur across several levels) are allowed for. Whether these forms of causation are actualised in real systems is according to him an empirical question. What makes Baas' contribution interesting here is that it allow us to see one approach in which the very idea of scientific explanation as a strictly deductive argument can be reinterpreted and explanations can be seen in a more dynamic and context-dependent setting, eventually themselves being emergent structures, "emergent explanations", and that this intuitive idea can be made precise and explicated formally. Questioning traditional notions of explanation may lead to a more general view of what constitute genuine scientific understanding of complicated phenomena. By paying attention to the general framework for description of higher-order structures (hyperstructures) which includes the mechanisms of observation, and which eventually allows for self-generation in such systems of new observational frames, new observers may also emerge in such a system (Baas, 1996).

It is a sensible intuition that the autonomy of biology in relation to physical science is grounded in the observational emergence of specific properties of biosystems, such as the self-reproduction of living cells. One may even understand the concept of life-as-an-autopoietic-system as the observational emergence in the physical space of systems that realize their own self-production, boundary and self-observation (through the boundary's distinctive or selective property, which is based on molecular recognition reactions by the membrane-bound proteins), and thus being `cognitive' in this primitive sense.[71] If this is true, an autopoietic system is autonomous because it realizes observational emergence of itself as an observer. This clearly has important semiotical implications and hints at a hidden connection between biosemiotics and the theory of autopoiesis.

It should be clear that the notion of emergence, as specified by the formal framework, does not in itself suffice for an ontodefinition of life. Rather, once a definition of a living system has been obtained, this definition implies that such a system is emergent.

A more overall inference from this discussion is that the increased interest for emergence in the sciences of complexity necessitates a deeper understanding of the nature of the modeling relation and the role of the observer in specifying the properties modeled and interpreting the resulting constructions.

Concluding remarks

If concepts of life as a general phenomenon are in use in biology, as argued above, can this be taken as a vindication of the principle of unity of knowledge? According to a universalist interpretation of the definition of life, the attempt to achieve general understanding in science cannot be dispensed with. Artificial Life was by its founders viewed as a contribution to reform and universalize theoretical biology to explain life in any kind, form, and medium, and to discover the general principles of evolution, adaptability, growth, development, behavior and learning. The analysis of the implicit very general definitions of life is here seen as supporting the claim that biology as a science of general processes of life should profit from interdisciplinarity and search for universal principles of organization. Artificial Life is simply a tool in this process, as mathematics and computer simulation is a tool in physics and chemistry. The origin of order in the universe and the emergence of biological organization on Earth and on other planets should be understood in a single (causal, historical, and physical) frame. The emergence of special principles of organization (e.g., a genetic code, and thus, of biological information) may grant biology conceptual autonomy, and may grant organisms a special ontology and mode of being -- but the evolution of the universe, life, and mind should ultimately be explained in a grand narrative provided by science and informed by semiotics and philosphy. Discovery of new laws of self-organization and evolution may eventually reform our picture of cosmos in a more `organic' direction, in which our perception of the world may be reenchanted. But we should not give up a search for single unified world picture in science.

A general concept of life is highly relevant in some contexts, e.g., in the contexts of discourses within general evolutionary biology, protobiology (research on the origin of life), artificial life, extraterrestrial life, philosophy of biology, and bioethics. A general concept of biological life may still vary with these different contexts. But we should not in general dismiss generality. Knowledge of the concrete must involve the specific as well as the universal. We might be tempted to perceive `the Disorder of Things' (cf. Dupré, 1993) as a sign of impossibility of general knowledge, but we should not resign to accept the disorder of thought. Unity of science in the positivistic sense is neither possible nor desirable, but post-modern abandonment of search for general principles will not do. Order as well as disorder is inherent in mind and nature.

This does not have to be in conflict with a sober critique of any form of scientism, because search for universality in science does not commit us to universalism in politics, religion or ethics. Furthermore, universality in science (as a value, or as a sign of modernity) is not the same as universality in politics or in culture.

Acknowledgements

This paper had not been possible without helpful discussions with colleagues and friends also interested in philosophy of nature, biology and signs. Thanks to all of them. The work was supported by The Faculty of Science, University of Copenhagen.

Notes

[1] This flavour of reductionism is indicated, for instance, by the romantic dictum of Spencer Brown: "To explain, literally to lay out in a plane where particulars can be readily seen. Thus to place or plan in flat land, sacrificing other dimensions for the sake of appearance. Thus to expound or put out at the cost of ignoring the reality or richness of what is so put out. Thus to take a view away from its prime reality or royality, or to gain knowledge and lose the kingdom." (G. Spencer Brown: Laws of Form. Here cited from Wilden, 1980, p.155).

[2] E.g., Elster, 1979. For a comprehensive account, see Salmon, 1990. Opinions differ on the specific meaning and implications of various types of explanations, but my aim here is more discuss new kinds of `emergent explanations' (see below) that to clarify the older typologies.

[3] Also called alife. Introductions in Langton, 1989; Emmeche, 1994a; Boden, ed., 1996.

[4] cf. Langton, 1989, p.22 ff. Langton seldom makes it clear that what is `seen' at both levels depends on our interpretation; he often writes as if only the GTYPE (his term) is `artificial' and interpretation-based while the PTYPE is a natural emergent real phenomenon, simply realized in the computational medium (cf. Emmeche, 1994b).

[5] cf.: Langton, 1989 p.39: "the interactions among the low-level entities give rise to the global level dynamics which, in turn, affects the lower levels by setting the local context within which each entity's rules are invoked. Thus, local behavior supports global dynamics, which shapes local context, which affects local behavior, which supports global dynamics, and so forth." Note, that even if this sounds a `downward causation' (which is ontologically problematic), it can be interpreted (in accordance with dynamical systems theory of attractors in the phase space) as not `effective' top-down causation but as a kind of `formative' causation in a more Aristotelian sense (Emmeche, Køppe and Stjernfelt, in prep.).

[6] As announced by Langton, 1989, and as shared by a majority within the field, as evident in the proceeding volumes published up to now. The rhetoric and research political negotiations around the establishment of AL as a scientific legitimate field has, as far as I know, not yet been analysed. A positive experience in participating in the Al meetings is an openness towards criticism and interest for the concerns of the sceptics among biologists, philosophers and the science studies people.

[7] Of course, there is no doubt that more modest interpretations of the goals of alife research can be given, for instance, simply to provide a more adequate parallel distributed information processing paradigm for modeling known biological processes, but the field as such was constituted by the far more ambitious ideas, boldly formulated by people as Chris Langton, Doyne Farmer, and Thomas Ray. I would be interesting to know what percentage of the alife community that would resort to a kind of "weak stance" if asked about their commitments to "strong artificial life".

[8] Thus the `break' with the reductionistic paradigm is not so deep as some of the alife proponents (e.g., Farmer and Belin 1992) seems to say. For instance, some of the logical positivists counted the eventual synthetic construction of life as one of the arguments for their working hypothesis of a (reductively) unified science, see Oppenheim & Putnam, 1958.

[9] In living systems, this amounts to what Jesper Hoffmeyer has called a code-duality between analogic and digital codes. See Hoffmeyer and Emmeche, 1991, Hoffmeyer 1997. Compare with what Pattee, 1977, has termed a `linguistic mode' of a complex system, in addition its dynamical (physical) mode. Though this additional intuition (c) refers to a biosemiotic property of living systems, neither biosemiotics nor any deep understanding of biology is necessarily involved in various proposals for the Artificial Life research programmes.

[10] This aspect of explanations has been emphasized in a biological context by epistemological constructivists such as Maturana & Varela, 1980, p.55: "An explanation is always a reproduction, either a concrete one through the synthesis of an equivalent physical system, or a conceptual one through a description from which emerges a system logically isomorphic to the original one, but never a reduction of one phenomenological domain into another."

[11] We shall not consider particular `operational' definitions of "dead" or "alive" organisms (e.g., fruit fly larva) that may be relevant in the particular experimental set-up.

[12] Molecular biologists live in an `epistemological culture' (in the sense of Knorr-Cetina, 1991) that is quite different from the culture of, e.g., high energy physicists. Universal theories seldom play a substantial role. Biology and its network of disciplines has a separate historically founded "moral economy" (Daston, 1995) - i.e., a set of norms and affect-saturated values for good science - that puts emphasis on quantification, facticity, empirical observations (as in experimental biology) and on description of novel and unique phenomena (especially in the natural history tradition of zoology, botany, ecology, biogeography, etc.).

[13] Even though definitions of life are missing, biology textbooks may give a kind of implicit definition by their expositions of for instance the molecular structure of living cells, giving the reader a general idea of life. E.g., the first six subheadings of chapter 1 in Alberts et al., 1983, are the following: "Simple biological molecules can form under prebiotic conditions"; "Polynucleotides are capable of directing their own synthesis"; "Self-replicating molecules undergo natural selection"; "Information flows from polynucleotides to polypeptides"; "Membranes defined [sic] the first cells"; "Mycoplasmas are the simplest living cells".

[14] Thus in Henderson's Dictionary of Biological Terms (Ninth Edition, Sandre Holmes, Longman, London, 1985) we meet "life cycle", "life form", "life zone" - all technical terms, but not "life". However, in The Penguin Dictionary of Biology (Eighth Edition, M. Abercrombie et al., eds., Penguin, London, 1992), we find a short clear definition (a variant of Maynard Smith's and "the replication definition" to be discussed below): "Life. Complex physico-chemical systems whose two main peculiarities are (1) storage and replication of molecular information in the form of nucleic acid, and (2) the presence of (or in viruses perhaps merely the potential for) enzyme catalysis."

[15] Mayr, 1982, p. 53. Mayr seems to imply that true definitions can only be given in the form of specifications of "the one and only" crucial defining property of the object to be defined, i.e., as a kind of the same essentialism, which he correctly criticizes in evolutionary biology.

[16] ibid., p. 53.

[17] ibid., p. 55. The computer metaphor of life is discussed in Oyama 1985; Emmeche and Hoffmeyer, 1991; Sarkar, 1996; Knudsen, 1996.

[18] as Mayr does in his further discussion, Mayr 1982: 59-67).

[19] We do not suppose that Mayr necessarily endorses SVDL in all details, we simply use Mayr's remarks to illustrate a common attitude among biologists; an attitude that is made explicit here as SVDL.

[20] This is why life, according to one interpretation of SVDL, may be seen as a `family-resemblance concept' (as suggested by Feldman, 1995). There may be a `fuzzy' borderline between living and non-living processes so that life is described as "a continuum property of organizational patterns, with some more or less alive than others" (Farmer & Belin, 1992, p. 819).

[21] A large minority of biologists would claim that it can be done in principle. This claim is often based on a refusal of the organicist ontology of separate levels of organization (see below), and by embracing a reductionist materialism as alternative.

[22] The word "foundation" should not be taken in the (foundationalist) sense of having secure and a priori foundations for scientific knowledge. On the autonomy of biological science, see Rosenberg 1985. Perceiving biology as an autonomous science is often related to the ontological notion of levels of organization. This widely accepted "organicistic" position holds that even though biology cannot be reduced to physics (as believed in the classical mechanicism in biology), one cannot ascribe any "hidden" mysterious qualities to living systems (as in vitalism). In 20th century biology, organicism has had many adherents, including J. H. Woodger, J. Needham, P. Weiss, C.H. Waddington, E. Mayr, R. Lewontin, R. Levins. Autonomy can also be conceived as an epistemological condition based on, e.g., an instrumentalist philosophy of science as in Rosenberg, 1994.

[23] see Haraway 1976, Sattler 1986. One can distinguish two kinds of organicism, 1) the reflective philosophical position of the biologists like Woodger and Weiss; 2) the informal standpoint of biologists expressed as belief in the reality of specific biological entities with emergent or relational properties irreducible to physics and chemistry, though the parts of the entities are recognized as chemical constituents (on emergent entities, see Blitz, 1992).

[24] Ruse 1995 (an entry in a companion volume to philosophy). The case of viruses as `borderline' between living (they have a genetic code, they can replicate) and `non-living (they have no metabolism of their own; they are transferred as inert crystal-like structures) is a classic and could be the subject of a separate paper. We will just notice that viruses as a (pathological) form of life presupposes (in the functional and evolutionary sense) the existence of living cells; thus they are better conceived of as pathological instances of life, a kind of ultimate parasites.

[25] Maynard Smith 1996.

[26] This formulation is nearly a reminiscence of Kant's discussion in Critique of Judgement [§ 66, second part]: "This principle, which is at the same time a definition, is as follows: "An organized product of nature is one in which every part is reciprocally purpose [end] and means. In it nothing is vain, without purpose, or to be ascribed to a blind mechanism of nature." (...) "it may be that in an animal body many parts can be conceived as concretions according to mere mechanical laws (as the hide, the bones, the hair). And yet the cause which brings together the required matter, modifies it, forms it, and puts it in its appropriate place, must always be judged of teleologically, so that here everything must be considered as organized, and everything again in a certain relation to the thing itself is an organ." (Kant 1790 [1951: 222]), even though Maynard Smith would give the traditional `mechanical' natural selection explanation of teleonomy. For discussions of darwinian teleonomy and Kantian teleology, see Mayr 1982 and Cornell 1986.

[27] Maynard Smith 1986, p. 7.

[28] This is not true for experiments concerning the origin of life, or biogenesis, where criteria for defining simple instances of autocatalytic macromolecular replication systems as living would be relevant.

[29] Such requirements or criteria for appropriateness of a definition will, of course, vary with the context of the use of a particular definition (as clearly emphasized by van der Steen, 1997), but I have chosen here to opt for a very general definition in the context of theoretical biology, not more specific contexts in particular areas of experimental research, as I think the biological concept of life is theoretically useful and already is more or less implicitly defined.

[30] A lot of work is done these days within cognitive philosophy of science, cognitive semantics, linguistics and literature theory on the cognitive role of metaphors for organizing conceptual schemes and other kinds of representations.

[31] I am here talking about the scientist's understanding of biological life, which is not necessarily the same as "the final reality" of life in a more phenomenological or existential sense - or just "social and mental life".

[32] The literature on the philosophy of biological taxonomy and on the species concepts are enormous, for an introduction, see for instance Mayr, 1987; chapter 6 in Sober, 1993.

[33] Lakoff and Johnson, 1980, p. 122. Some elements of their view of definition may extend, in a modified way, to science, namely that also scientists use concepts, that "are not defined in an isolated fashion, but rather in terms of their roles in" [scientific, theoretical and experimental] "natural kinds of experiences" (p.125); so one should investigate how scientists "get a handle on the concept - how they understand it and function in terms of it" (p.116).

[34] I am not so sure, but it is my general impression that few philosophers of science has really investigated the various kinds and levels of definition. Quine is an important exception.

[35] On the distinction between transformational (as in ontogenesis) and variational change (as in phylogenesis), see Lewontin, 1982; Levins and Lewontin, 1985, chapter 3.

[36] And perhaps only one-way transfer of digital information (non-Larmarckian inheritance, cf. Maynard Smith 1996).

[37] Dawkins joined the Artificial Life programme with enthusiasm, cf. Dawkins 1989.

[38] Even though evolutionary epistemologists like Popper, Lorenz and Riedl (cf. Campbell, 1974) have suggested that for instance scientific development proceeds like a process of natural selection; the specific mechanisms of change can only in a metaphorical sense be like the blind variational mechanisms at work in nature.

[39] As critically discussed by Pattee, 1989, Emmeche, 1992.

[40] See Emmeche 1992, 1994a.

[41] Kim, 1995. See also Berckermann, Flohr & Kim 1992. The standard notion of emergence does not specify if prediction or reduction is impossible in principle or in practice, or in what non-trivial senses one cannot predict such properties, or if the "arising" of such properties are a consequence of descriptive or physical processes, etc. Many philosophers prefer supervenience for emergence as a concept characterising the dependence relation between entities or properties at different levels.

[42] Here I assume that supervenience indicates emergence, but the concept of supervenience may not adequately capture all we want to express with the proposition that something is emergent. Emergence seems to demand a supervenience relation of some kind. E.g., that the combinatorial properties of the genetic code are emergent may imply that the code is not simply supervenient on the physicochemical properties of a single organism, but on those properties plus their causal history (Macdonald, 1995).

[43] This second aspect is usually expressed quite informally, e.g., in Dawkins, 1989, p.201: "The process of emergence [of biomorphs] was to be evolution by the Darwinian process of random mutation followed by nonrandom survival".

[44] It is important to note (a) that a short exposition cannot do the semantically complicated and reflexively abstruse writings of Maturana and Varela any justice, so the reader must go to the original papers for further study; (b) we shall only deal with the definition of life aspect of the theory here, not its epistemology or ontology; and (c) though Maturana and Varela have cooperated, one finds some variations in their respective views on autopoiesis and autonomy, so our remarks in this brief note are fairly preliminary. An introduction to the theories are given by Mingers, 1989. For those who wish a more formal and logical introduction to autopoiesis, and the more general concept of autonomy, the treatise of Varela, 1979, is a must.

[45] This means that whatever takes place in living systems as living systems, takes place "as necessarily and constitutively determined in relation to themselves because their being defined as unities through self-reference" (Maturana and Varela, 1980, p. xiii).

[46] Ibid., p. 135. The notion that living systems is autopoietical systems "in the physical space" implies the possibility of having autopoietic systems in abstract, non-physical spaces; a problem in the definition as a definition of life that has been discussed in detail by Fleischaker, 1988.

[47] This has implications for the origin of autopoiesis, which "cannot be a gradual process" (Maturana and Varela, 1980, p. 94). "We can describe a system and talk about it as if it were a system which, with a little transformation, would become an autopoietic system because we can imagine different systems with which we compare it, but such a system would be intermediate only in our description, and in no organizational sense would it be a transition system" (ibid.).

[48] Compare Maturana and Varela, 1987; Maturana 1978. A critique of this aspect of the theory is given by Mingers, 1990. The application of the concept of autopoiesis in Varela's work on the immune system will not be considered here.

[49] This complex is sometimes criticized for being theoretically inadequate, e.g., Sarkar, 1996; Stuart, 1985.

[50] Fernandez, Moreno, and Etxeberria, 1991; Moreno, Umerez, and Fernandez, 1994, Moreno, Etxeberria and Umerez 1995. This `San Sebastian approach' to theoretical biology and philosophy of biology may provide a stepping stone for an integrated theory of life's thermodynamic, informational-semiotical, and autopoietic aspects.

[51] Hoffmeyer 1997; Hoffmeyer, forthcoming; see also Sebeok & Umiker-Sebeok, eds., 1992. Introductions to biosemiotics on the internet: http://www.gypsymoth.ento.vt.edu/~sharov/biosem/welcome.html

[52] Though I have attempted to discuss them a little further in a Danish introduction to biosemiotics (Emmeche, 1997) where it is regarded as an alternative philosophy of nature.

[53] including the ecosystem as the semiotic `bio-culture'; compare the concept of `semiosphere' (Hoffmeyer, 1997).

[54] I take the term `original meaning' from Haugeland, 1985. Though he uses it in a the context of AI (how to get meaning from pure syntax), the problem described here (how to get meaning from pure physics) is similar.

[55] This is often advanced as an argument against the notion of downward causation. E.g., if a mental state by downward causation changes a neural state in the brain, then we could see this as violating the neuro-physical causal closure of the system on the micro level where one physical state (that corresponds to the neural state) should be a sufficient cause for the next neural and physical state without further intervention from `non-physical' (mental) causes (cf. Kim, 1993). However, one of the problems in this discussion is to make a clear demarcation of "system" and "state".

[56] This was what Niels Bohr emphasized: "When we attempt a unified description of which we ourselves are part we met with the problem of a wholeness where the observational standpoint is lost. (...) Only when a section [line] is drawn between a part of it and the rest, the idea of observation can be obtained." (Bohr, MSS, No.21, 19.8.1954, here cited from p. 100 in Favrholdt, 1994). Bohr believed that committing oneself to claims that "everything existing is material", or "everything existing is mental" (or semiotical for that matter!), one has committed a philosophical failure of presupposing `The Angelic Point of View".

[57] See also Emmeche, Køppe and Stjernfelt (in prep.), and the work of Jaegwon Kim.

[58] The prime example of this position is Roger Sperry's interactionism. In contrast to the strong version, medium downward causation does not involve the idea of a strict "efficient" temporal causality from an independent higher level to a lower one, rather, the entities at various levels may enter part-whole relations (e.g., mental phenomena control their component neural and biophysical sub-elements). The control of the part by the whole can be seen as a kind of functional (teleological) causation, which is based on efficient, material as well as formal causation in a multinested system of constraints. (The kind of determinative relation between part and whole is not quite clear, and the term "interaction" is not the best for the kind of relationship envisaged). Thus, "Mind is conceived to move matter in the brain and to govern, rule, and direct neural and chemical events without interacting with the components at the component level, just as an organism may move and govern the time-space course of its atoms and tissues without interacting with them" (Sperry 1987).

[59] For further details, see Kampis 1991; Emmeche 1994a; Moreno, Etxeberria & Umerez 1995.

[60] Nils A. Baas (personal communication). Recently there have been some thoughts on the possibility of higher order emergence in computational models of life in which the emergent patterns may change the very dynamical rules of the system, not yet conclusive. See Fontane, Wagner & Buss 1994.

[61] Certainly "organism plus environment" may in some contexts be a more accurate term for a biosystem, as the organism-environment relation is crucial one (cf. Lewontin, 1982). I do not intent to claim that the organism is a privileged entity in contrast to cell lines (compare Buss, 1987) or to other levels of selection. The controversy over levels of selection within the philosophy of evolutionary biology is not in conflict with the general notion of life discussed here.

[62] This is evident in Stuart Kauffman's work, see the commentary by Burian & Richardson, 1996.

[63] This was Peirce's view, which may be relevant for the emergence of the first cells during the protobiological-biological transition (ca. 4.6 billion years ago). With respect to viruses these are often mentioned as a borderline case (in the cell they are alive, in crystal form, they are not), e.g. in the statement cited above from Ruse, 1995. This does not make sense: Even though viruses can be crystalline, are much simpler than a cell, and have no autonomous metabolism, they cannot be fully understood if one does not comprehend their `biology', their way to parasite a living cell when they are produced, or if one fail to understand that viruses themselves are products of evolution. In this sense, a virus is no mysterious borderline case, but a genuine biological phenomenon, an ultimate parasite of a truly biological origin.

[64] Historical accounts are to be found in Blitz, 1992; Berckermann, Flohr and J. Kim, eds., 1992.

[65] The answer these days to concerns for compatibility with a scientific world view seems to be "who cares?", because of a more widespread accept of relativism and perspectivism. In this very superficial sense of the Zeitgeist, emergence of irreducible new properties that constitute their own level of explanation or phenomenology is in harmony with the post-modern outlook.

[66] The asserted isomorphy between causal processes in Nature and algorithmic, syntactic processes in alife (and other dynamical) models faces fundamental problems as indicated by the work of Robert Rosen (e.g., 1985, 1988) and Kampis.

[67] It has been argued by Newman, 1996, that being in the basin of a strange (chaotic) attractor is an emergent property of any chaotic nonlinear dynamical system which makes emergence a very general phenomenon in (models of) the physical world.

[68] though arguments from algorithmic information theory can be given that some patterns in fact are, cf. Dennett 1991.

[69] The role of the observer and the interpretative nature of bottom-up emergent explanations has been emphasized in various ways (which cannot be discussed in detail here) by Pattee, 1989; Cariani, 1990; Kampis, 1991; Emmeche, 1994a.

[70] We shall not go into the technical details, just give an informal hint about the basic idea (the following notation is simplified and must not be taken as Baas' original one).

[71] Cognitive in the sense of Maturana and Varela where a cognitive domain is the entire domain of all interactions in which an autopoietic system (an organism) can enter without loss if identity. I think that Baas' framework allow for an interpretation of the theory of autopoiesis that adequately integrates evolution and emergence of autonomy.

References

Alberts, B., Bray, D., Lewis, J. , Raff, M., Roberts, K., Watson, J. D., 1983. Molecular Biology of the Cell. New York: Garland Publ.

Baas, N.A., 1994: `Emergence, Hierarchies, and Hyperstructures', p.515-537 in: C. G. Langton, ed., Alife III, Santa Fe Studies in the Sciences of Complexity, Proc. Volume XVII, Redwood City, CA: Addison-Wesley.

Baas, N.A., 1996: `A framework for higher order cognition and consciousness', p. 633-648 in S. Hameroff, A. Kaszniak and A. Scott, eds.: Towards a science of consciousness. Cambridge, Mass.: The MIT Press.

Blitz, D., 1992. Emergent Evolution. Dordrecht: Kluwer.

Boden, M.A. (ed.) 1996: The Philosophy of Artificial Life. Oxford: Oxford University Press.

Burian, R.M. & R.C. Richardson, 1996: "Form and order in evolutionary biology", pp. 146-172 in Boden, ed. 1996.

Buss, L. W., 1987: The Evolution of Individuality. Princeton, N.J.: Princeton University Press.

Campbell, D. T. 1974: `Evolutionary epistemology'. pp.413-463 in: P.A. Schilpp, ed.: The Philosophy of Karl Popper. Book 1. La Salle, Ill.: Open Court [also in Plotkin, H.C., ed. 1982. Learning, Development, and Culture. Essays in evolutionary epistemology. Chichester and New York: John Wiley & Sons.]

Cariani, P., 1990: `Emergence and Artificial Life', pp. 775-797 in: Artificial Life II. Santa Fe Institute Studies in the Sciences of Complexity Proceedings Vol. X, C. G. Langton, C. Taylor, J. D. Farmer and S. Rasmussen, eds. Redwood City, Calif.: Addison-Wesley.

Coleman, W. 1971. Biology in the Nineteenth Century: Problems of Form, Function, and Transformation. London: John Wiley.

Cornell, J.F. 1986. `Newton of the Grassblade? Darwin and the problem of organic teleology.' Isis 77: 405-421.

Berckermann, A., H. Flohr and J. Kim, eds., 1992: Emergence or Reduction? Essays on the Prospects of Nonreductive Physicalism. Berlin & New York: Walter de Gruyter.

Dawkins, R. 1976. The Selfish Gene. New York and Oxford: Oxford University Press.

Dawkins, R. 1989. `The evolution of evolvability', p. 201-220 in Langton, ed., Artificial Life (Santa Fe Institute Studies in the Sciences of Complexity, Vol. VI), Redwood City, California: Addison-Wesley.

Daston, L. 1995. `The moral economy of science'. Osiris 10: 3-24.

Dennett, D.C., 1991: "Real patterns", The Journal of Philosophy 88(1): 27-51

Dupré, J., 1993: The Disorder of Things. Metaphysical Foundations of the Disunity of Science. Cambridge, Massachusetts: Harvard University Press.

Elster, J. 1979: Forklaring og dialektikk. Oslo: Pax Forlag.

Emmeche, C. 1992a: `Modeling Life: A note on the semiotics of emergence and computation in artificial and natural systems', pp. 77-99 in: Biosemiotics. The Semiotic Web 1991, Thomas A. Sebeok & Jean Umiker-Sebeok, eds., Mouton de Gruyter Publishers, Berlin and New York.

Emmeche, C. 1992b. `Life as an abstract phenomenon: is Artificial Life possible?' pp. 466-474 in: F. J. Varela and P. Bourgine, eds.: Toward a Practice of Autonomous Systems. Proceedings of the First European Conference on Artificial Life. Cambridge, Mass.: The MIT Press.

Emmeche, C. 1994a. The garden in the machine. The emerging science of artificial life. Princeton University Press.

Emmeche, C. 1994b. `The computational notion of life'. Theoria - Segunda Época 9(21): 1-30. (San Sebastian).

Emmeche, C. 1997. `Den biosemiotiske tanke'. [The idea of biosemiotics (in Danish)]. p. 62-94 in: Keld Gall Jørgensen (ed.): Anvendt Semiotik. Gyldendal, København.

Emmeche & Hoffmeyer, J. 1991. `From language to nature - the semiotic metaphor in biology', Semiotica 84 (1/2), 1-42.

Emmeche, C., Stjernfelt, F. & Køppe, S. (in prep.): "Levels, emergence and three versions of Downward Causation", chapter contributed to Peter Bøgh Andersen, Niels Ole Finnemann, Peder Voetmann Christiansen & C. Emmeche, eds.: Downward Causation. (manus. version 1997; book expected to be published by Aarhus University Press 1999)

Engelsted, N. 1989. `What is the Psyche and How did it Get Into the World?' In: Essays in General Psychology, edited by N. Engelsted, L. Hem and J. Mammen, 13-48. Aarhus: Aarhus University Press.

Favrholdt, D., 1994: Fysik, bevidsthed, liv. Studier i Niels Bohrs filosofi. Odense Universitetsforlag, Odense.

Farmer, J.D. and Belin, A. d'A. 1992. `Artificial Life: The coming Evolution.' pp. 815-838 in: Artificial Life II. Santa Fe Institute Studies in the Sciences of Complexity Proceedings Vol. X, C. G. Langton, C. Taylor, J. D. Farmer and S. Rasmussen, eds. Redwood City, Calif.: Addison-Wesley.

Feldman, F. 1995. `Life', p. 272-74 in: J. Kim and E. Sosa, eds. A Companion to Mataphysics. Oxford: Blackwell.

Fleischaker, G. R. 1988. `Autopoiesis: the status of its system logic'. BioSystems 22: 37-49.

Fernandez, J., Moreno, A. and Etxeberria, A. 1991. `Life as emergence: the roots of a new paradigm in theoretical biology'. World Futures. The Journal of General Evolution 32, 133-149.

Fontana, W., Wagner, G. and Buss, L.W., 1994: `Beyond digital naturalism', Artificial Life 1: 221-227.

Haraway, D. J. 1976. Crystals, Fabrics, and Fields: Metaphors of Organicism in Twentieth-Century Developmental Biology. New Haven: Yale University Press.

Hoffmeyer, J., 1997: Signs of Meaning in the Universe. Bloomington: Indiana University Press.

Hoffmeyer, J. (forthcoming). `Semiotic aspects of biology: Biosemiotics', in: R. Posner, K. Robering and T. A. Sebeok, eds. Semiotics: A Handbook of the Sign-Theoretic Foundations of Nature and Culture. Berlin & New York: Mouton de Gruyter.

Hoffmeyer, J. & C. Emmeche, 1991: "Code-Duality and the Semiotics of Nature". p. 117-166 in: On Semiotic Modeling, edited by Myrdene Anderson & Floyd Merrell. Berlin: Mouton de Gruyter.

Haugeland, J., 1985: Artificial Intelligence: the very idea. The MIT Press, Cambridge, Mass.

Hull, D. L. 1981. `Units of evolution: a metaphysical essay' pp.23-44 in: U.J. Jensen and R. Harré, eds.: The Philosophy of Evolution. New York: St. Martin's Press.

Kampis, G. 1991. Self-modifying Systems in Biology and Cognitive Science. New York: Pergamon Press.

Kampis, G., and Csányi, V. 1987. `Replication in abstract and natural systems.' BioSystems 20: 143-152.

Kant, I. 1790. Kritik der Urteilskraft (Critique of Judgement, translated by J. H. Bernard, New York: Hafner Publishing Company. 1951).

Kim, J., 1993: Supervenience and Mind, Cambridge: Cambridge University Press.

Kim, J. 1995: "Emergent properties", p.224 in: T. Honderich, ed.: The Oxford Companion to Philosophy. Oxford: Oxford University Press.

Kincaid, H., 1988: "Supervenience and explanation", Synthese 77: 251-281

Knorr-Cetina, K. 1991. `Epistemic cultures: Forms of reason in science', History of Political Economy 23: 105-122.

Knudsen, S. 1996: By the grace of gods - and years and years of evolution. Analysis of the development of metaphor in scientific discourse. Ph.D.-thesis, Dept. of Languages and Culture, University of Roskilde.

Küppers, B.-O. 1992: `Understanding complexity', p. 241-256 in Beckermann, Flohr & Kim, eds.: Emergence or Reduction? Essays on the Prospects of Nonreductive Physicalism. Berlin & New York: Walter de Gruyter.

Lakoff, G. and Johnson, M. 1980. Metaphors we Live By. Chicago and London: The University of Chicago Press.

Langton, C.G. 1989: "Artificial life," pp.1-47 in: Langton, ed., Artificial Life (Santa Fe Institute Studies in the Sciences of Complexity, Vol. VI), Redwood City, California: Addison-Wesley. (an updated version in: M.Boden, ed., 1996).

Levins, R. and Lewontin, R. 1985. The Dialectical Biologist. Cambridge, Massachusetts: Harvard University Press.

Lewontin, R.C. 1982. `Organism and environment'. pp. 151-170 in H. C. Plotkin, ed.: Learning, Development, and Culture. Chichester: John Wiley & Sons.

Macdonald, C. 1995: `Psychophysical supervenience, dependency, and reduction' p. 140-168 in E.E. Savellos & Ü.D.Yalçin, eds.: Supervenience. New Essays. Cambridge, University of Cambridge Press.

Maturana, H. R. 1978. `Biology of language: the epistemology of reality.' pp. 27- 63 in: G.A. Miller & E. Lenneberg, eds. Psychology and biology of language and thought: Essays in honour of Eric Lenneberg. New York: Academic Press.

Maturana, H.R. and Varela, F.J. 1980. Autopoiesis and Cognition. The Realization of the Living. (= Boston Studies in the Philosophy of Science Vol. 42). Dordrecht: D.Reidel.

Maturana, H.R. and Varela, F.J. 1987. The Tree of Knowledge - The Biological Roots of Human Understanding. Boston: New Science Library, Shambhala.

Maynard Smith, J. 1986. The Problems of Biology. Oxford: Oxford University Press.

Maynard Smith, J. 1996. `Evolution--natural and artificial'. p. 173-178 in Boden, ed., 1996.

Mayr, E. 1982. The Growth of Biological Thought. Diversity, Evolution, and Inheritance. Harvard University, Cambridge: The Belknap Press, .

Mayr, E. 1987. `The ontological status of species `. Biology and Philosophy 2: 145-166 (reprinted in Mayr. 1988. Toward a new philosophy of biology. Cambridge, Massachusetts: Harvard University Press).

Mingers, J. 1989. `An introduction to autopoiesis - implications and applications'. Systems Practice 2(2): 159-180.

Mingers, J. 1990. `The philosophical implications of Maturana's cognitive theories'. Systems Practice 3(6): 569-584.

Moreno, A., Umerez, J., and Fernandez, J. 1994. `Definition of life and the research program in artificial life.' Ludus Vitalis 2 (3): 15- 33.

Moreno, A., Etxeberria, A., and Umerez, J. 1995: `Universality without matter?', pp. 406-410 in R. Brooks & P. Maes, eds. : Artificial Life IV. The Santa Fe Institute. Redwood City, California: Addison-Wesley.

Newman, D.V., 1996: `Emergence and strange attractors', Philosophy of Science 63, pp. 245-261.

Oppenheim, P. & H. Putnam, 1958: "Unity of Science as a Working Hypothesis", pp.3-36 in H. Feigl, M. Scriven & G. Maxwell, eds.: Minnesota Studies in the Philosophy of Science: Vol. II. Minneapolis, MN: Minnesota University Press. (reprinted in: R. Boyd, P. Gasper & J.D. Trout, eds., 1991: The Philosophy of Science. Cambridge, Massachusetts: The MIT Press).

Oyama, S., 1985: The Ontogeny of Information. Cambridge: Cambridge University Press.

Pattee, H.H., 1977: `Dynamic and linguistic modes of complex systems'. Int.J.General Systems 3: 259-266.

Pattee, H.H. 1989. `Simulations, realizations, and theories of life.' pp. 63-77 in: Langton, ed.: Artificial Life (= Santa Fe Institute Studies in the Sciences of Complexity, Vol. VI). Redwood City, Calif.: Addison-Wesley. (also in M. Boden, ed., 1996).

Peirce, C. S. 1955. Philosophical Writings of Peirce. (ed. Justus Buchler). New York: Dover Publications.

Rosen, R. 1985: `Organisms as causal systems which are not mechanisms: an essay into the nature of complexity', pp.165-203 in R. Rosen (ed.): Theoretical Biology and Complexity. Orlando, Oxford: Academic Press.

Rosen, R., 1988: `Effective processes and natural law', pp. 523-537 in: R. Herken, ed.: The Universal Turing Machine. A Half-Century Survey. Oxford: Oxford University Press.

Rosenberg, A. 1985. The Structure of Biological Science. Cambridge: Cambridge University Press.

Rosenberg, A. 1994. Instrumental Biology or the disunity of science. Chicago and London: The University of Chicago Press.

Ruse, M. 1995. `Life'. In: T. Honderich, ed.: The Oxford Companion to Philosophy. Oxford: Oxford University Press.

Sagan, C. 1973. `Life.' In: The Encyclopaedia Britannica, 15th ed., Macropaedia vol. 10. London: William Benton.

Salmon, W.C. (1990): Four decades of scientific explanation. Minneapolis: University of Minnesota Press.

Sarkar, S. 1996: `Biological Information: A Skeptical Look at Some Central Dogmas of Molecular Biology'. in: S. Sarkar, ed.: The Philosophy and History of Molecular Biology: New Perspectives. Dordrecht: Kluwer. (forthcoming).

Sattler, R. 1986. Biophilosophy. Berlin: Springer-Verlag.

Sebeok, T.A. and J. Umiker-Sebeok (eds.), 1992. Biosemiotics: The Semiotic Web 1991. Berlin: Mouton de Gruyter.

Sober, E. 1993. Philosophy of biology. Oxford: Oxford University Press.

Sperry, R.W., 1987: "Consciousness and causality", p.164-166 in R. L. Gregory, ed.: The Oxford Companion to the Mind. Oxford University Press, Oxford

Stuart, C. I. J. M. 1985. `Bio-informationel equivalence.' Journal of theoretical Biology 113: 611-636.

van der Steen, W. J., 1997: `"Life" as a Concept: The Elusiveness of Ultimate Meaning', Ultimate Reality and Meaning (forthcoming in August 1997).

Varela, F.J. 1979. Principles of Biological Autonomy. New York: Elsevier North-Holland.

Wilden, A. (1980): System and Structure. Essays in Communication and Exchange. 2nd ed. New York: Tavistock Publications.

Wittgenstein, L. 1972 [1953]. Philosophical Investigations. Oxford: Basil Blackwell.

*