The way in which we define "life" has important implications for ethics, for the definition of "animal" and even for the range of creatures that can be credited with having minds. The definition which I am seeking here is both a conceptual one which does justice to our fundamental intuitions regarding life, and a theoretical one that attempts to describe just what it is that makes something alive. Beginning with Aristotle's "four-dimensional" account of life, I trace the history of philosophical thinking about the term "life" and locate the key positions taken on a "conceptual map". I argue that the recent invention of so-called "artificial life", which can be defined as an attempt to re-create the basic principles governing living systems, within a computational medium (Digital Life Lab at CalTech, 2002) has generated a philosophical crisis regarding the definition of "life". What were formerly thought to be the defining features of living things can be mimicked by abstract computational systems which are governed by very simple rules (Wolfram, 2002, pp. 824 - 825).
After reviewing the various definitions of "life" in the literature, I conclude that the teleological (end-directed) properties of life are basic and incapable of being reduced to lower-level properties. However, I also maintain that these properties supervene upon lower-level properties, allowing scientists to identify entities which are alive, according to strict empirical criteria. This does not entail reductionism, as the language we use to describe these properties cannot be divorced from teleology. I then attempt to construct a list of sufficient and necessary conditions for something's being alive, and address the question of why does the distinction between living and non-living matters ethically.
The question of whether the life versus non-life distinction is a fundamental one has practical consequences regarding both the nature of "mind" and the scope of our ethical concerns.
Thanks to recent advances in robotics, scientists are close to being able to construct robots whose behavioural feats equal or surpass those of many animals in sophistication - such as the InsBot robot, which is capable of infiltrating a group of cockroaches, influencing them and altering their behaviour (Pest Control Times, 16 November 2004, Web address http://www.pctonline.com/news/news.asp?ID=3071). The behavioural feats of AIBO, the robotic dog, are also impressive. These developments have important implications for the way we think about animal minds. There is a good deal of popular support for the view that if all of an animal's behavioural feats can be duplicated by a human-built robot or computer, then there is no need to impute a mind to the animal - which is why many people, upon hearing about some feat of animal cognition, are apt to object: "Yes, but even a computer could do that." However, if there turns out to be some fundamental difference between robots and animals, then we can no longer infer that an animal whose behaviour can be duplicated by a robot has the same mental abilities as the robot. The simple fact that the robot is not alive might constitute a valid reason for calling its behaviour mindless.
The scope of our ethical concerns would also be dramatically affected if there turned out to be no basic differences between living things and artifacts. We could choose to narrow our scope and deny interests and intrinsic value to those animals whose feats could be duplicated by a machine, or we might decide that machines matter too, as Davison (1999) proposes.
It is not my intention to provide a detailed historical account of philosophical and scientific answers to the question of "What is Life?" in this thesis (but see Weber (2003) for a useful summary and critical evaluation of 20th century attempts to define "life"). Instead, what I propose to do is to situate some of the answers that have been given to this question, on a kind of "conceptual map".
Aristotle's account of life
Aristotle's four categories of causality - efficient, material, formal and final - provide a useful starting point for constructing such a conceptual map. Aristotle originally proposed these categories as an explanation of change: the efficient cause is that by which the change is brought about (i.e. the agent of change), the material cause is that in which the change is brought about (i.e. the subject of the change), the formal cause is that which is brought about (i.e. the transformation in the entity undergoing change), and the final cause is that for which the change is brought about, in those cases where the change is a goal-directed or purposeful one. Contemporary English usage of the term "cause" is much narrower than Aristotle's: today, the term is generally reserved for the first of his four causes.
Insofar as Aristotle taught that all four categories of causality were needed to answer the question, "What is it that makes (or causes) something to be alive?", his explanation of life can be characterised as four-dimensional. As Cameron (2000) convincingly demonstrates, there is overwhelming textual evidence that Aristotle regarded the four dimensions as mutually irreducible, and that he viewed final causality as a sui generis ontological category.
Aristotle criticised his materialist contemporaries, among them Democritus, for neglecting to include final and formal causality in their account of life, and for attempting to explain the phenomena of life using only two dimensions of causality - the material and efficient causes (Generation of Animals v.1 778b7-10; v.8 789b1-15).
The Scientific Revolution, commencing around 1600, also provided a two-dimensional, reductionist account of scientific explanation: every phenomenon in the universe was held to be explicable in terms of mechanical laws governing the motion and collision of particles of matter (Wilson, 1996; Wikipedia, 2004, arts. "Scientific Revolution", "Mechanism", "Vitalism").
Mechanism and Vitalism: Was Aristotle a Vitalist?
The philosophical debate about life from the 17th through to the early 20th century was dominated by a clash between mechanists and vitalists. Unfortunately, the philosophical discussion of this controversy is obscured by a lack of agreement in the literature (reviewed by Cameron, 2000, pp. 33-40) regarding the definition of the terms "mechanism" and "vitalism".
Aristotle has sometimes portrayed as a proto-vitalist. I would argue that this characterisation is erroneous, for two reasons. Firstly, Aristotle's four-dimensional account of life was much richer and far more sophisticated than the accounts that were subsequently advanced by scientists in either of the two "camps". Secondly, although Aristotle rejected reductionism, he regarded living things as continuous with the rest of nature. Overall, his views were closer to those of the mechanists than to the doctrines of the vitalists (Cameron, 2000).
One fact that emerges from Wilson's (1996) overview of the historical debate between mechanists and vitalists regarding "life" is that both camps argued within a two-dimensional framework. The key points at issue were: what kinds of laws applied to living things, and what sort of material they were made of. Mechanists asserted that the known laws of physics and chemistry, operating on ordinary matter in motion, could explain the phenomenon of life, while vitalists typically envisaged living things as being controlled by a "vital force" (elan vital) that obeyed special non-physicochemical laws and directed the continual generation of a unique class of organic compounds found within the bodies of living organisms. Indeed, vitalistic chemists (such as Berzelius) believed that these compounds could only be generated within living organisms, until other chemists succeeded in synthesising them via non-biological processes in the 19th century. These advances, coupled with experimental demonstrations that living things obeyed the laws of mechanics and thermodynamics, led to the eventual demise of vitalism as a scientific theory (Wilson, 1996). In short, the vitalistic explanation of life was quasi-mechanical from its inception, unlike Aristotle's four-dimensional account.
Additionally, there is no evidence that Aristotle thought living things violated physical laws, but there is abundant textual evidence (Cameron, 2000, pp. 41 ff.) that he regarded living things as materially continuous with non-living things - he believed in the spontaneous generation of life, for instance.
The philosophical discussion of mechanism and vitalism has been helpfully clarified by Cameron (2000), who distinguishes between strong and weak versions of both schools of thought (Table 1.1).
Table 1.1 - Strong and Weak Versions of Mechanism and Vitalism, as defined by Cameron (2000, pp. 36-39) (italics mine)
Minimal definition: the belief that living things obey the laws of physics and chemistry.
Minimal definition: the belief that the "aliveness" of an organism is completely independent of the interactions between its parts.
|Strong mechanism: Organisms are nothing but chemical machines - all their properties and parts are wholly reductively identifiable with and explicable in terms of the laws of physics and chemistry as they apply outside of living systems. (Cameron has adapted this definition from Mayr, 1982.)||Strong vitalism: That property or entity in virtue of which living things are alive is:
(i) irreducibly distinct from its physical properties and parts,
(ii) to some extent determines the course of physical events in living things, and
(iii) breaks or overrides physical laws in the course of (ii).
|Weak mechanism: Organisms are 'something more' than chemical machines; although their activities do not conflict with the laws of physics and chemistry, at least some of their properties or parts are not wholly reductively identifiable with or explicable in terms of the laws of physics and chemistry as they apply outside of living systems (adapted from Mayr, 1982).||Weak vitalism: That property or entity in virtue of which living things are alive is (i) irreducibly distinct from its physical properties and parts, (ii) to some extent determines the course of physical events in living things, and (iii) fails to depend [i.e. does not completely depend - V.T.] for its existence on the interactions of such physical parts of living things as there may be.|
It should be noted that according to the broad definition proposed by Cameron, mechanism is not tied to a two-dimensional account of life. Indeed, Cameron (2000, p. 40) classifies Aristotle as a weak mechanist. (Wilson (1996), on the other hand, defines mechanism as a reductionist program for biology, which is why he classifies Aristotle as an organicist rather than a mechanist.)
Vitalism is no longer scientifically tenable: advances in the field of biology have decisively refuted even the weaker vitalistic claim that "living things somehow fail to depend on their physical parts for their being alive" (Cameron, 2000, p. 37). On the other hand, strong mechanism is regarded by modern scientists as too simplistic: contemporary mechanists "do not accept... that animals are 'nothing but' machines... The phenomena of life have a much broader scope than the relatively simple phenomena dealt with by physics and chemistry" (Mayr, 1982, p. 52).
That leaves us with weak mechanism. However, Cameron (2000, p. 35) faults Mayr's (1982) definition of weak mechanism for being too broad: as it stands, it is fully compatible even with Cartesian dualism, as a dualist need not maintain that non-physical minds violate the laws of physics and chemistry in their interactions with bodies. Since the prevailing scientific mind-set rejects all forms of dualism except some of the weaker varieties of property dualism, Cameron (2000, p. 38) has therefore defined a position - which he labels "sophisticated mechanism" - as a "precisification" of weak mechanism. Cameron characterises Aristotle's "emergentist" view of life as a via media between sophisticated mechanism and vitalism.
Table 1.2 - Sophisticated Materialism and Emergentism, as defined by Cameron (2000, pp. 36-39) (italics mine)
(i) living things are wholly composed of physical parts;
(ii) some parts or properties of living things are not reductively identifiable with underlying physical parts and properties but are reductively explicable in those terms; and
(iii) no property not reductively identifiable with underlying physical parts or properties has 'its own' causal powers; the causal powers of organisms are exhausted by the causal powers of their physical parts.
(i) living things are wholly composed of physical parts;
(ii) all parts and properties of living things are causally or logically dependent for their existence on the interactions of their physical microparts;
(iii) not every part or property of living things is a part or property fully reductively identifiable with the parts and relations of those same materials outside of living things; and
(iv) the 'emergent' parts or properties noted in (iii) exert 'their own' causal influence over the course of their microparts' careers without breaking laws governing the behavior of microparts and properties.
Table 1.3 - Aristotle's emergentist position compared and contrasted with sophisticated materialism and vitalism, according to Cameron (2000, pp. 36-40)
|School of thought||Sophisticated Mechanism||Aristotle's Emergentism||Weak Vitalism|
|Are there any activities in living things which violate physical laws?||No||No||No|
|Upward determination: Do higher level properties and parts of organisms depend for their existence on micro properties and parts?||Yes||Yes||No|
|Downward causation: do living things have higher-level properties whose causal powers are not exhausted by the causal powers of their underlying parts?||No||Yes||Yes|
Because Aristotle accepts upward determination, Cameron classifies him as a mechanist. According to Cameron, the crucial difference between Aristotle and contemporary mechanists is the issue of downward causation, which characterises Aristotle's emergentism (2000, p. 39).
The return of form
I would suggest that there are two important differences between 20th century scientific accounts of life and 19th century mechanist accounts: a revival of scientific interest in Aristotle's dimension of form, and the creation of a new, temporal dimension for explaining the phenomenon of life.
Hopkins' influential address to the British Association for the Advancement of Science in 1913, in which he set forth his vision of the emerging field of biochemistry, could be said to mark the renewal of scientific interest in form (Weber, 2003). What Hopkins proposed was a new model of the cell as a chemical machine, whose internal organisation was just as important as the chemical processes that regulated it.
The publication of Schroedinger's influential book "What is Life?" (1944) reflects a similar pre-occupation with form. Specifically, Schroedinger hypothesised that the molecular basis of life was an "aperiodic" solid with a "miniature code" embedded within its structure. His work had a powerful influence on Crick and Watson, the co-discoverers of DNA.
More recently, Davies (1999) has claimed that "[t]he secret of life lies, not in its chemical basis, but in the logical and informational rules it exploits" (1999, p. 212). Although his willingness to characterise life in terms of metabolism and reproduction makes him sound like a scientific mechanist, his recognition of the fact that both of these mundane processes require organisms to embody vast amounts of information clearly places him in the "formalist" camp.
Life's fifth dimension
Meanwhile, 19th century physicists investigating thermodynamics had developed a fifth dimension for explaining life: the temporal dimension. Whereas for Aristotle, "the phenomenon ... most basic in the apparent flux of the world was the unity and persistence of the individual living being" (Smith, 1976, p. 62, cited in Weber, 2003), modern scientists came to regard flux as a defining feature of life, as it responds to an ever-changing environment. Two recurring features of life - its order and its ability to evolve - attest to this fact in different ways.
According to the modern conception, the very stability of an organism is an indicator of its life-long active struggle against a hostile environment that is trying to break it down. Some physicists even attempted to define living things by their ability to continually generate order from disorder through their metabolism (Wolfram, 2002, p. 1178). Schroedinger's (1944) contribution to the debate was his proposal that living things achieved this task through the thermodynamics of open systems far from equilibrium.
Schroedinger's account of life can thus be seen as one which invokes the dimensions of form and the arrow of time to characterise living things in terms of both their structural and thermodynamic properties. The same could be said for Kauffman's (1993, 2000) recent proposal that a fourth law of thermodynamics might be needed to account for the self-organising properties of life.
The ability to evolve is such a pervasive feature of life that some neo-Darwinian biologists (e.g. Dawkins, 1976, 1986) are prepared to call anything that can evolve "alive". Their evolutionary paradigm of life relies heavily on the explanatory dimensions of form (i.e. DNA) and time (which produces variation through sexual recombination and mutation). Both evolutionary and thermodynamic accounts of life are inherently temporal, but there is an important difference between them: in thermodynamic accounts, the changes envisaged are not those occurring within individual organisms during their life-long struggle against the forces of chaos (entropy), whereas in evolutionary accounts, slow, subtle inter-generational genetic alterations drive evolution, enabling some species to successfully adapt to changes occurring in their environment, over long periods of time.
The view I defend in this chapter is that it is possible to formulate a rigorous definition of life, but that we need five dimensions to do so: Aristotle's four causes, plus time. My neo-Aristotelian account of what it means to be alive explicitly invokes teleological causality, like that of Cameron (2000); unlike Cameron, I claim that there is nothing particularly mysterious about final causality, and that it supervenes upon Aristotle's other dimensions of causality - especially form - but is not reducible to these other dimensions. Lastly, I attempt to show that Aristotle's notion that each species of living organism possesses its own nature is fully compatible with Darwin's assertion that species change over time.
In this section, I argue that the quest for a definition of life has generated a full-blown philosophical crisis which has been largely ignored until very recently. "Crisis" is not a word I use lightly. The fact that all of the functional definitions that purported to list necessary and sufficient conditions for being alive have fallen foul of counter-examples does not constitute a crisis - after all, there are many things that cannot be defined in such a rigorous fashion. Not even the philosophical failure of the weaker cluster definitions of life that have replaced our old, simple definitions merits the term "crisis". Rather, what has caused the crisis is recent research with abstract computational systems, showing that even systems obeying very simple rules can instantiate properties once regarded as hallmarks of life (Wolfram, 2002). These systems are not generally considered alive or even life-like. To make matters worse, these simple systems can be found in nature too: indeed, they pervade the natural world. Commonly used definitions of life thus turn out to be full of holes, making ordinary distinctions between "animate" and "inanimate" appear to be nothing more than folk categories.
1.A.3.1 Problems associated with the quest for single-attribute definitions of life
Traditionally, philosophers and scientists have approached the problem of defining "life" by looking for some distinguishing attribute or hallmark - for instance, some special capacity, function or mode of behaviour. However, "single-attribute" definitions like these are notoriously susceptible to three kinds of problems: counter-examples, borderline cases and category challenges.
Functional definitions of life illustrate how vulnerable single-attribute definitions are to counter-examples, as the following table shows.
Table 1.4 Hallmarks of life that are shared by non-living things
|Feature||Non-living things that instantiate this feature|
|Spontaneous self-movement||Any machine with primitive sensors
(see Descartes, 1968, Discourse 5, pp. 73-74; Wolfram, 2002, p. 823).
|Responsiveness to stimuli||As above|
|Nutrition, defined by Wolfram as "the ability to take disorganized material and spontaneously organize it" (2002, p. 824).
Equivalent to a thermodynamic definition of life, insofar as the exercise of this function results in a local entropy decrease.
|"[A]ll sorts of systems - from crystals to flames - also do this" (Wolfram, 2002, p. 824).
(Viruses, on the other hand, cannot generate or store energy, but have to derive their energy, and all other metabolic functions, from their host cell.)
|Tendency toward homeostasis,
(i.e. an equilibrium of parameters that define the internal environment of an organism).
|Crystals can achieve equilibrium too (Cleland, 2002).|
|Reproduction||As J.B.S. Haldane (1947, p. 56) pointed out, if we accept that life is essentially a pattern of chemical processes, it follows that any pattern that begets a pattern similar to itself - as a flame does, for instance - can be described as reproducing.
(Note: For Haldane, one thing that distinguished an organism from a flame was its additional capacity for self-regulation. However, "self-regulation" is a rather vague concept, and as we shall see below, it occurs in purely mechanical as well as living systems.)
Other problems associated with this hallmark of life: is a mule alive? What about a fertile individual, who is unable to reproduce in the absence of a fertile individual of the opposite sex?
|Mechanism of heredity||Crystals in nature instantiate a form of heredity: they may be perfectly aligned until a specific point is reached, in which a flaw has arisen. This flaw has a tendency to percolate down the subsequent layers of crystal, setting up a rudimentary system of heredity.|
|Evolution by natural selection||Evolution by natural selection turns out to be a common feature of inanimate natural systems such as clay crystallites and autocatalytic networks of chemical species (Bedau, 1996, p. 338).|
Of course, the fact that scientists have failed to find a single attribute that defines "life" does not mean that they cannot; it may simply mean (as I shall argue) that we need a better definition. But even if no definition could be found that was free of embarrassing counter-instances, the project of defining life would not have to be abandoned. One could still attempt to formulate looser definitions which listed properties that living things almost invariably possess and that inanimate objects generally lack. I discuss these cluster definitions below.
It is commonly argued that borderline cases, such as viruses, evolving RNA strings in molecular genetics laboratories, frozen organisms and the problem of pinpointing the exact point in time when the first true life-form appeared on the primordial earth (Bedau, 1996, p. 333; 1998, p. 132), render problematic any attempt to define necessary and sufficient conditions for life. Cameron (2000, pp. 30-32) responds to these criticisms by arguing that the demand for a clearcut definition is dogmatic and question-begging: it assumes that an adequate definition of X must provide us with an objective method for deciding, in all cases, if something is X. Imposing this kind of "decision procedure requirement" on all definitions imposes an improper epistemological constraint on natural categories such as "living things", which may not turn out to have sharp boundaries. Instead Cameron proposes a "no-epistemological baggage thesis" (2000, pp. 31-32), which states that an adequate definition of life "must tell us what conditions a thing must satisfy in order to be alive", but "need not yield objective decision procedures for determining in hard cases whether or not something is alive". The definition of "life" is thus compatible with a limited degree of ambiguity - a point recognised by Aristotle himself, who appears not to have regarded "being alive" as an "on-off" property (Cameron, pp. 40-42).
The project of defining hallmarks of living things has also come under fire from philosophers and scientists who question its implicit assumption that "life" is a property of individual organisms. Lovelock's Gaia hypothesis claims that the entire chemical and biological environment around the surface of the earth (including the oceans and the atmosphere) is alive. According to another bold proposal by Bedau (1996, p. 336), "life" is primarily a property of systems that are capable of "supple adaptation" - that is, "responding appropriately in an indefinite variety of ways to an unpredictable variety of contingencies". Individual organisms that make up such systems are said to be alive only in a derivative sense.
Cameron (2000, pp. 23-27) argues that while these new proposals may prove to be scientifically illuminating, they fail as definitions of life, insofar as they fail to satisfy a reasonable constraint which applies to all proper definitions - even to theoretical definitions which attempt to revise our concept of what something is. The constraint Cameron proposes is a categorial one: any proper definition of a concept must respect its fundamental ontological categories. The number 2 cannot be re-defined as a concrete material object; and neither can "life" be re-defined as a property of systems rather than individuals. If it should prove impossible to define life on the individual level, then it would be better to abandon the term "life" altogether.
I conclude that the foregoing arguments do not appear to place the enterprise of defining life in serious danger. Even if we cannot find a single defining feature of life, cluster definitions might still do the trick.
1.A.3.2 Can cluster definitions work?
There is a widespread feeling among philosophers and biologists that the search for such a strict definition of life in terms of some single "hallmark" property, such as reproduction, is doomed to failure (Cameron, 2000, pp. 11-12). Many experts despair of being able to do anything more precise than writing down "a list of properties that we associate with life" (Farmer and Belin, 1992, p. 818). Bedau clarifies the thinking behind the cluster definitions of life proposed by these authors:
The individual properties in the cluster are held to be typically but not necessarily possessed by living entities; the diversity of living phenomena are thought to bear only a Wittgensteinian family resemblance (Bedau, 1996, p. 335).
Cluster definitions of life tend to fall into two groups - short and long - both of which have been subjected to criticism.
1.A.3.2(a) Short cluster definitions
Table 1.5 - Properties listed in some recent short cluster definitions of life
|A potentially self-perpetuating open system of linked organic reactions, catalysed by stepwise and almost isothermally by complex and specific organic catalysts [enzymes] which are themselves produced by the system (J. Perret, 1952).|
|"Teleonomic" or purposeful behavior; autonomous morphogenesis; and reproductive invariance (Monod, 1971; cited in Bedau, 1996, p. 336).|
|Self-reproduction, genetics and evolution; and metabolization (Crick, 1981; cited in Bedau, 1996, p. 336).|
|Metabolism; self-reproduction; mutability (Kuppers, 1985; cited in Bedau, 1996, p. 336).|
|We regard as alive any population of entities which has the properties of multiplication, heredity and variation (J. Maynard Smith, 1975).|
|Metabolism; having parts with functions (J. Maynard Smith, 1986; cited in Bedau, 1996, p. 336).|
|Self-reproduction; and the capacity for open-ended evolution (Ray 1992; cited in Bedau, 1996, p. 336).|
The following definitions listed in the on-line "Life in the Universe" educational outreach program (organised by CERN, the ESA the ESO) were given by a panel of experts interviewed by Newton magazine (June 2001, Italian version) in a report entitled "Segnali di Vita Aliena" (Signs of Alien Life):
Haboku Nakamura - Biology Institute, Konan University, Kobe, Japan:
Andre Brack - Centre for Molecular Biophysics of CNRS, France:
Camilo J. Cela-Conde - Dept. of Philosophy, University of Baleares:
Short cluster definitions of life are vulnerable to two flaws pertaining to the definition of any natural category.
First, although the above definitions share several common themes (reproduction, metabolism, heredity, variation, evolution and autocatalysis), the attributes cited lack unity, and give us no reason to think that the term "life" is a bona fide natural category:
A cluster offers no explanation of why that particular cluster of properties is a fundamental and ubiquitous natural phenomenon... (Bedau, 1996, p. 336).
Here, Bedau seems to be appealing to a general definitional requirement for any natural property, which I shall refer to as the unity-of-criteria condition:
Any adequate definition of a natural property P should explain the unity of the criteria for something's being P: what is it that makes all of them (and only them) criteria for being P?In this case, P refers to the natural property of being alive. An adequate definition of this property should explain why the assorted criteria listed, and no others, are criteria for life.
The foregoing cluster definitions fall foul of another definitional problem: they fail to account for what Cameron (2000) calls the problem of unity:
The problem of unity (PU): A condition of adequacy on an adequate definition of a property, P, is that it explain the unity of the items which fall into the extension of P: what is it that makes them all P? (2000, p. 50, italics mine).Here, Cameron presents his adequacy condition as one which applies to any property P. This is somewhat too sweeping - Wittgenstein's counter-example of the property of being a game comes to mind. However, the property of being alive purports to define membership of a natural category (unlike the category of all games). The unity of the objects belonging to a natural category is logically prior to any human declaration that they have something in common; for artificial categories, it is the other way round. Cameron's unity requirement is a reasonable one, provided it is restricted to any property P that defines membership of a natural category. It follows that any adequate definition of life should tell us why the term "life" applies to the range of things we call alive, and excludes inanimate objects.
Whereas the unity-of-criteria condition strives to unify the criteria for instantiating a natural property, Cameron's condition pertains to the unity of the entities instantiating it. The above cluster definitions of life fail to satisfy either requirement.
A final problem with short cluster definitions is their vulnerability to robust counter-examples. For instance, physicists have recently created cell-like blobs of gaseous plasma which, despite their lack of a mechanism for heredity, satisfy the four main criteria generally used to define living cells: they possess a boundary layer that separates their interiors from the external environment; they can replicate by splitting in two; they can grow by taking in atoms and splitting them into ions and electrons to replenish their boundary layers, and they can communicate information by emitting electromagnetic energy that causes neighbouring "cells" to resonate (Cohen, 2003). However, nobody would think of describing these gaseous blobs as "alive". It seems that a more complete description of life is needed.
1.A.3.2(b) Long cluster definitions
Because they are comprehensive, long cluster definitions are much less vulnerable to the counter-examples cited above. Long definitions have the additional merit of deepening our understanding of "life". A representative sample of such definitions is listed in the table below.
Table 1.6 - Properties listed in some recent long cluster definitions of life
Mayr, 1982, p. 818
(Cited in Bedau, 1996, p. 336.)
1. All levels of living systems have an enormously complex and adaptive organization.
2. Living organisms are composed of a chemically unique set of macromolecules.
3. The important phenomena in living systems are predominantly qualitative, not quantitative.
4. All levels of living systems consist of highly variable groups of unique individuals.
5. All organisms possess historically evolved genetic programs which enable them to engage in 'teleonomic' processes and activities. [See below for definition of 'teleonomic'. - V.T.]
6. Classes of living organisms are defined by historical connections of common descent.
7. Organisms are the product of natural selection.
8. Biological processes are especially unpredictable.
The term "teleonomy" is defined by Maturana and Varela (1979) as follows:
the element of apparent purpose or possession of a project in the organization of living systems, without implying any vitalistic connotations. Frequently considered as a necessary if not sufficient defining feature of the living organization.
Farmer and Belin, 1992, p. 818
(Cited in Bedau, 1996, p. 335.)
information storage of self-representation;
functional interactions with the environment;
interdependence of parts;
stability under perturbations; and
the ability to evolve.
G. Gordon Winder, 1993.
(C. Gordon Winder is Professor Emeritus in the Department of Earth Sciences at the University of Western Ontario.)
Life is a succession of energy-producing electro-chemical processes
by a naturally occurring, simple or complex organism composed of a combination of molecules,
forming cells, which consume 'food' and produce 'waste', both consisting of solid, aqueous, and gaseous matter;
the process is called metabolism;
the organism is capable of living within the environment without dependency on any other organism;
energy use is manifest by growth with size limits for most;
self-replication with each offspring slightly different;
capable of modifying their living environment, both beneficially and detrimentally;
with eventual termination of energy production, or death.
Exceptions are egg, sperm, spore, seeds and virus, which do not consume food or produce waste; the first four are replication structures, and the fifth has premature life-terminating capabilities.
(Definition proposed at a 1993 symposium entitled "What is Life? Define Life", organised by C. Gordon Winder at the biennial International Society of History, Philosophy, Social Studies of Biology (ISHPSSB) conference at Brandeis University, Waltham, Massachusetts.
Sidney Fox - South Alabama University, USA:
Living beings are protein-made bodies formed by one or more cells that communicate with the environment through information transfer carried out by electric impulses or chemical substances, and capable of morphological evolution and metabolism, growth and reproduction.
The above definition is listed in the on-line "Life in the Universe" educational outreach program (organised by CERN, the ESA the ESO). Fox was one of a group of experts interviewed by Newton magazine (June 2001, Italian version) in a report entitled "Segnali di Vita Aliena" (Signs of Alien Life).
Shane Sarver, 1999
Sarver is an Associate Professor of Biology at Black Hills State University, South Dakota. Sarver lists the following seven features of living things on his Web page "Definition of Life and Intro. to Zoology":
(1) chemical uniqueness (living things contain large complex macromolecules, made up of the same building blocks - nucleic acids such as DNA and RNA, proteins, lipids and carbohydrates);
(2) hierarchical organisation (molecules are nested into cells, the building blocks of all living systems, which are nested into tissues, tissues into organs, and organs into an organism);
(3) reproduction, which enables the transmission of traits from parents to offspring;
(4) a genetic program, or script (usually contained in DNA), that codes for the make-up of living things;
(5) metabolism (extraction of chemical energy from nutrients);
(6) the occurrence of characteristic phases of development in a life cycle; and
(7) interaction with the environment: ecology is the study of this interaction.
(N.B. The definition of life on Sarver's (1999) Web page is also used by other universities. It can be accessed from the University of Toronto Zoology department home page by clicking on the link "Careers in Zoology", which takes you to the Web page, "What is Zoology?". This page has a link, "Definition of Life", to Sarver's Web page.)
Former "Science" editor-in-chief Daniel Koshland (2002) lists seven "pillars of life":
(1) a program, i.e. "an organized plan that describes both the ingredients themselves and the kinetics of the interactions among ingredients as the living system persists through time" (Koshland, 2002, p. 2215);
(2) improvisation, or a way of changing its master program (achieved on Earth through mutation);
(3) compartmentalisation (a surface membrane or skin, and for large organisms, a subdivision into cells, in order to preserve the ingredients required for chemical reactions at their required concentrations);
(4) energy (which on Earth comes from the Sun or the Earth's internal heat), to keep living systems metabolising;
(5) regeneration (this includes reproduction), to compensate for the wear and tear on a living system;
(6) behavioural adaptability to environmental hazards; and
(7) seclusion, or some way of preventing one set of chemical reactions from interfering with another, in a cell.
One could fault most of these definitions for their inclusion of properties that are either too ad hoc (why should the presence of DNA, proteins or even carbon figure in our definition of "life"?) or too vague and uninformative (e.g. "information storage" or "interaction with the environment"). Another pertinent criticism of these criticisms is that as Cleland and Chyba (2002) argue, a laundry list is a dangerous way to define natural categories. The two authors cite the example of water: one could attempt to define "water" as anything that is wet, odorless, tasteless, and thirst-quenching, but even this definition "could still allow substances that superficially resemble water to be incorrectly classified as water".
Perhaps the greatest failing of these definitions is that they only serve to exacerbate the philosophical problems with regard to the definition of life, which they were meant to resolve. Bedau's comment on Mayr's "especially comprehensive" cluster definition is illuminating:
Mayr's list ... cannot help but deepen our sense of wonder and perplexity about what root cause could conspire to make this striking collection of features present in such an indefinite diversity of natural phenomena. We want an account of why these properties all coexist. Rather than settling this question, the list raises it (Bedau, 1996, p. 336, italics mine).
It is readily apparent that long cluster definitions of life, like short ones, fail to satisfy the unity-of-criteria condition or to solve Cameron's problem-of-unity (2000, p. 50).
Bedau concludes that "[a] cluster conception is a fall-back position that can be justified only after all candidate unified views [of life] have failed" (1996, p. 336).
Despite these failings, the long cluster definitions listed above are not without merit, as they embody some rich philosophical insights, which I intend to make use of below when putting forward my own neo-Aristotelian account of life. I also argue that some of these definitions, rightly interpreted, contain empirical criteria that enable us to distinguish living from non-living things.
1.A.3.3 Is it too soon to formulate a good definition?
In the face of the above criticisms, many scientists, whom I shall refer to as "inopportunists", regard the question of life's definition as unanswerable for the time being. One commonly cited reason for this pessimism is that all the life-forms we are familiar with are terrestrial and share a common origin (Cleland, 2002). Despite its apparent diversity, it is claimed, life on Earth is really only a single case. While I am prepared to grant for the sake of argument that all life on Earth is monophyletic, I do not regard this line of argumentation as particularly convincing. Terrestrial bias would certainly be a relevant factor if we were trying to identify the chemicals necessary for life, but the problem with defining life lies elsewhere. "Life" is not a purely chemical category: if it were, there could be no debate about whether so-called "artificial life-forms" such as computer viruses are truly alive. In ordinary usage, the term "life" is usually taken to designate some kind of activity or process.
Another argument against constructing preliminary definitions of life is that as yet, we lack a general "theory of life". Cleland and Chyba (2002) remark that medieval alchemists classified many kinds of substances as "water", including nitric acid (which was called aqua fortis). Only with the advent of molecular theory were scientists able to understand why nitric acid, despite sharing many of the properties of water, is nonetheless not water. But as McKay (2004) remarks, the analogy of life with water is not a particularly good one: the term "water" designates a material, whereas "life" refers to some kind of process. In any case, the chemical processes that take place in living cells are well-understood.
One could argue that even if we understand the chemistry of life, we do not yet understand how blind chemical processes could have given rise to the vast quantity of information that must have been embodied even in the first living cell, in order for it to be able to reproduce and metabolise (Davies, 1999). While I have no quarrel with Davies' suggestion that an understanding of how life started may shed light on what it is, I would like to point out that the latter question is logically independent of the former. Generally, we do not have to know how X's originate in order to grasp what they are.
I conclude that the "inopportunists" have failed to make a good case against pressing ahead with the attempt to define life.
1.A.3.4 The new crisis regarding the definition of life
Although the ad hoc cluster definitions of life we examined above proved to be philosophically unsatisfying, they are hardly a cause for panic. Cluster definitions, after all, possess the virtue of rendering the definitional problems caused by awkward borderline cases more manageable, while keeping the phenomenon of life reasonably well-delineated.
Recently, however, Stephen Wolfram (2002) has argued that no "general definition of life, independent of the details of life on earth ... can actually be given", because "almost any general feature that one might think of as characterizing life will actually occur, even in many systems with very simple rules" (2002, p. 825, italics mine). The originality of Wolfram's contribution consists not in his assertion that all of the alleged distinguishing characteristics of life are shared by some inanimate objects, but that a very large class of simple systems - which nobody would even think of calling alive - share these characteristics. The definitions of life turn out to be as full of holes as a leaky sieve: they are not even approximately correct. This situation, I would argue, merits the term "philosophical crisis".
Table 1.7 Hallmarks of life that are shared even by very simple systems
|Feature||Non-living things that instantiate this feature|
|Metabolism - "the ability to take disorganized material and spontaneously organize it" (Wolfram, 2002, p. 824).||"Self-organization is actually extremely common even among systems with simple rules", including "all sorts of systems" in nature (Wolfram, 2002, p. 824).|
|Self-reproduction||"[I]n the 1950s abstract computational systems were constructed that also had this ability. Yet it seemed that they needed highly complex rules - not unlike those found in simple cells. But in fact ... even systems ... with remarkably simply rules can still manage to show self-reproduction" (Wolfram, 2002, p. 824, italics mine).|
|Evolution by natural selection||This is a common feature of man-made computational systems. Lenski et al. (2003) have recently created digital organisms - simple computer programs that self-replicate, mutate, compete and "evolve" solutions to complex problems in a Darwinian, step-wise fashion, demonstrating that "complex functions can originate by random mutation and natural selection" (Lenski et al., 2003, p. 139; see also Knight, 2003).|
|Unpredictable behaviour||Wolfram's work on computational irreducibility (2002, pp. 737-750) demonstrates that unpredictable behaviour is a common feature of the inanimate world. For all except the very simplest systems that are found in nature, there are no predictive short-cuts: even with a complete knowledge of the rules and initial conditions of a system, it can still take an irreducible amount of computer work to calculate its future behaviour, which means that "in effect there can be no way to predict how the system will behave except by going through almost as many steps of computation as the evolution of the system itself" (2002, p. 739).|
|Complexity||A vast range of systems, even "ones with very simple underlying rules ... can generate at least as much complexity as we see in the components of typical living systems" (2002, pp. 824-825). This claim is elaborated in Wolfram's Principle of Computational Equivalence.|
In a nutshell, Wolfram's Principle of Computational Equivalence says that "there is essentially just one highest level of computational sophistication, and this is achieved by almost all processes that do not seem obviously simple" (2002, p. 717). More precisely: (i) almost all systems, except those whose behaviour is not "obviously simple", can be used to perform computations of equivalent sophistication to those of a universal Turing machine, and (ii) it is impossible to construct a system that can carry out more sophisticated computations than a universal Turing machine (2002, pp. 720 - 721; the latter part of the Principle is also known as Church's Thesis).
Wolfram suggests that the reason why we can easily distinguish living from non-living systems in everyday life is that living things share a lot of specific structural and chemical features - such as "being made of gelatinous materials and having components [like] proteins, enzymes, cell membranes and so on - and ... being based on specific chemical substances ... water, sugars, ATP and DNA" - reflecting "their long history of biological evolution" (2002, p. 825). Untutored laypeople may not grasp chemical concepts (such as "gelatinous material") at the micro-level, but they can certainly recognise chemical properties at the macro-level. We all know a bone when we see one. Hence our ability to recognise members of the folk category "life".
But if chemistry alone is what makes something alive, then one has to query the scientific and ethical relevance of the term "life".
Incidentally, Wolfram's Principle of Computational Equivalence has a very important philosophical consequence for the debate about how living things differ from computers. What it means is that we can no longer classify living things in a separate category from computers: living things - bacteria, plants and animals (including human beings)- are computers, insofar as their bodies can be used to perform the kinds of computations that a Turing machine can perform. The question of whether some computers qualify as living things presupposes that computers are a subset of the set of living systems, whereas in fact, the reverse is the case. Philosophically, it would be more appropriate to ask whether there is a significant difference between artificial computational devices and living organisms.
1.A.3.5 What is wrong with the foregoing definitions of life?
I suggest that there are two major problems associated with the above definitions of life, and I propose to characterise these problems in terms of the five "dimensions" of life which I highlighted earlier (Aristotle's four causes plus time). The two problems are a general (though by no means universal) failure of scientists to adequately characterise the formal features of living things, coupled with an almost complete neglect of the teleological aspect of life.
The formal features of life identified in the preceding definitions are deficient on various counts. Some of the formal attributes listed were arbitrarily specific, focusing on particular structures such as the DNA molecule. Yet even if it turned out that all life was DNA-based, we would still need to know what made DNA special.
Other accounts were too vague, attempting to characterise life in terms of general features that can be readily found even in the inanimate world (such as complexity, or a self-replicating pattern of chemical processes), or which establish only a quantitative difference between living and non-living things (such as information content).
Some accounts (Monod, 1971; Winder, 1993; Sarver, 1999) contained references to the distinctive morphological features of living things - such as cells, tissues and organs. However, the instructions needed to assemble the structures that make up the body of a living creature are contained in its genetic program. This suggests that the program that directs the make-up of these structures should be considered the central formal feature of living organisms.
A few of the long cluster definitions (Mayr, 1982; Sarver, 1999; Koshland, 2002) actually highlighted the existence of a genetic program that codes for the make-up of living things. But "programs" alone cannot suffice to define living things: programs (defined broadly as rules that govern the behaviour of an entity) are ubiquitous in nature (the laws of physics are a well-known instance), as well as in abstract computational systems (Wolfram, 2002, p. 383). What makes the programs in living things fundamentally different from other programs? Any satisfactory account of the form of an organism has to answer this question.
Our provisional conclusion is that the prospects of being able to define life in terms of some combination of efficient causality (which features prominently in functional definitions of life), matter (invoked in chemical as well as functional definitions), form (which figures in many long cluster definitions) and time (thermodynamic and evolutionary definitions) appear to be remote. These factors alone do not suffice to determine that something is alive: as Wolfram (2002) has shown, most of the features instantiating these dimensions are actually very common in nature.
Aristotle's dimension of final causality has been curiously neglected in most of the contemporary definitions of life given above, with a few notable exceptions: Mayr's (1982) fifth condition refers to teleonomic processes and activities in organisms (made possible by their genetic program), Monod's (1971) definition also refers to "teleonomic" or purposeful behaviour; Maynard Smith (1986) makes having parts with functions a defining feature of life; and Sarver's (1999) second condition - hierarchical organisation - suggests some kind of teleological organisation. It would be tempting to invoke this much-neglected dimension as the solution to our definitional problems regarding "life", but before we can do so, we need to address the reasons why many philosophers and scientists have chosen to set aside finalistic (teleological) accounts of life.
1.A.3.6 Is teleology redundant?
Final causation is out of favour in modern scientific circles, as it is incompatible with a "scientific world-view that countenances only efficient causation" (Buller, 1999, p. 6, italics mine). Accordingly, those who wish to do away with teleological talk in science have two choices: eliminate the very concept of teleology, or reduce it to something more scientifically respectable.
Cameron (2000, 2004) has mounted a sustained critique of these attempts to dispense with teleology. I review and assess the merits of Cameron's arguments in the Appendix to Part A of this chapter. The eliminativist option, as it turns out, is not a viable one. Teleology is real. Proponents of final causation, such as Cameron (2000, 2004), have therefore focused their attack on reductionism, arguing that attempts to reduce teleological properties to structural properties (the systems approach to teleology) or historical properties (the etiological account of functions) fail to account for their goal-directedness and biological normativity. The etiological account is currently a popular one in scientific and philosophical circles. In order to assessing the merits of the etiological account in the light of Cameron's (2000, 2004) criticisms, I have selected a particularly clear and far-sighted exposition of this account by one of its philosophical proponents (Varner, 1998). What is striking about Varner's etiological account is that it not only attempts to explain the functionality of organs in etiological terms, but the very distinction between life and non-life, in etiological terms. However, I conclude that Varner's account appears to be at odds with the available evidence, and that it fails to reduce teleology to the dimensions of efficient causality and time.
While I endorse the bulk of Cameron's criticisms of scientific attempts to do away with teleology, I part company with him on a key point: I accept that the structural and/or historical properties of some part of an organism can indeed determine whether it is goal-directed and subject to biological norms. However, I shall argue in section 1.B below that it is impossible in principle to describe the formal and historical properties of these parts without employing some teleological language. Hence the attempt to reduce teleology to some other explanatory "dimension" of life is doomed to failure. I conclude that teleology remains an irreducible, ontologically primitive fact of life.
1.A.3.7 Are there any good methodological reasons for rejecting teleological accounts of life?
Despite the problems facing non-teleological accounts of life, there are many philosophers and scientists who reject teleological explanations for purely methodological reasons. Cameron (2000) addresses some common reasons for excluding final causality from scientific explanations, refuting oft-repeated assertions that teleology implies a pre-existing commitment to animism, or panpsychism, or the existence of God, or the benevolence of nature, or a belief in "progressive" evolution, or vitalism (2000, pp. 215-219). In fact, teleology is compatible with the rejection of any or all of these beliefs. A related worry is that belief in teleology commits one to backward causation, but this objection seems to mistakenly presuppose that teleological causation is simply efficient causation running backwards in time, when in fact, final causality is not the same as efficient causation of any sort (2000, p. 218). Backward causation, as understood by Aristotle, simply means that some biological processes occur for the sake of future benefits to the organism. I discuss downward and backward causation at further length in section 1.B.4.
It is frequently objected that teleological explanations are vacuous, but Cameron (2000, pp. 219-220) argues that they are no more vacuous than explanations invoking efficient causation. The idea of something's being directed towards an end may seem obscure, but the notion of something's possessing a causal power to bring about a certain result is surely equally mysterious. If teleology is an irreducible part of the natural world, he argues, we should put aside our metaphysical prejudices and accept it.
1.A.3.8 Is a teleological account of life intelligible?
More specifically, Cameron (2000, pp. 240-279) proposes that teleology is a strongly emergent property of organisms, in the sense that although it depends causally for its existence on the interactions between their parts, it is irreducibly distinct in kind from the properties of these parts, and possesses causal powers not possessed by the parts and their interactions.
Emergence may seem a strange, almost magical notion to some, but Cameron contends that if we accept Hume's argument that "we are in a natural state of ignorance with regard to the powers and influence of all objects" when we consider them a priori, it follows that "there are no a priori bars on what types of events, properties, or entities might emerge from the causal interactions of complex groupings of micro-entities" (2000, p. 228). Nor can there be any bars on downward causation: for all we know, higher-level states such as final causes may be able to affect and even direct micro-level states. (Citing Arthur Lovejoy (1927), Cameron suggests that those who oppose emergentism do so because they unconsciously subscribe to a medieval "preformationist assumption" about causality: the doctrine that an effect is not understood until and unless "the eye of reason could somehow discern it in the cause" (2000, p. 228).)
Although I believe that Cameron has done a commendable job of arguing that it is metaphysical prejudice, rather than sound methodology, which underlies the scientific dismissal of teleological explanation, I would suggest that his account of teleological explanation leaves two major questions unanswered. First, there is the empirical question: exactly how can scientists tell which entities possess a telos or final cause? What kind of observations would a scientist need to make to confirm its presence?
Second, if final causes are an ontologically irreducible given, then why do we always look for non-finalistic explanations of biological mishaps such as death, injury and malformation? Why do material, formal or efficient causal explanations of these events seem satisfying while teleological explanations leave us dissatisfied? For instance, we can explain a person's death in: (a) efficient causal terms ("She died as a result of a lightning strike"); (b) material terms ("She died from a bacterial infection"); (c) formal terms ("She died of a heart attack"); or even (d) temporal, thermodynamic terms ("She died of old age - or more precisely, the ends [telomeres] of the chromosomes in her body cells became so short as a result of accumulated wear-and-tear that her body cells could no longer replicate themselves"). Any of these explanations of death sounds perfectly all right, but the teleological explanation "She died because her body was no longer capable of functioning as it should, in accordance with its telos" leaves us wanting to ask: "Yes, but why was it no longer capable of doing so?" The foregoing examples suggest that teleological explanations, even if irreducible, need to be supplemented with lower-level explanations.