Against units in biology

The living world, it seems to me, causes no end of trouble for those who would classify it. Its levels, ranks, hierarchies and units all seem to be clear enough, until we encounter troublesome cases. Then they get very troublesome indeed. So I want to say, there are no ranks or set units in biology a priori, and very few and limited, a posterori.

Suppose we take one instance: the "units of selection debate" that was so widely discussed after the publication of a number of seminal works in the 1960s and 1970s. Genes were held to be what evolved, against group selectionists who thought populations were the units (or a unit, depending on how enthusiastic you were) of evolution and selection.

This debate still rages nearly 50 years after it began. Why? Is it that we have two (or more) mutually inconsistent theoretical approaches? Or is it that there just is no fact of the matter? Maybe anything that meets certain criteria are "the" units of selection. In other words, if, in a particular group of organisms, such things as Xs are selected, then in that group they are a unit of selection. But there may be other things that are also selected. Some of these may include or depend upon Xs, but they need not. So, gene centrism seems to have evaporated somewhat in the past decade or so in favour of "multilevel selection theory". As one school of thought (Developmental Systems Theory) has it, genes are like the Triumvirate which included Julius Caesar, who was "first among equals", prima inter pares. This is the "parity thesis", according to which all resources of a cell or an organism, including DNA, are subjectible to selection; this doesn't imply that all resurces continually are, of course, for unless they meet the hereditability condition necessary for selection, and they vary in ways that allow some degree of novelty, and there are fitness differences between these variations, they aren't selected one way or the other. These are the necessary preconditions for a selection process: heredity, variety, differentiability.

So we ought to stop looking for a single level of selection. What counts as a level of selection depends on what the particular organismic lineage bears as its hereditary units (does it also have prions, for instance?), and whether or not they vary more or less than other hereditary units (if genes can vary but prions cannot, genes will be selected more effectively that prions will).

This is called pluralism, according to its critics, and some of its proponents. I think this is a misnomer. Pluralism has overtones of anarchism, constructivism, relativism. For many philosophers of science, including me, these things are neither all that is true about science nor virtues in and of themselves. Some degree of freedom in science is essential for the proper functioning of science, but there are opposed "forces" both in favour of latitude and against it. If we simply did accept the Fullerene view of science as being whatever some science power elite said it was, science would have nothing more to recommend it than astrology. But I digress. The point is that pluralism is the right way to approach a topic that is, actually, plural. But one should be no more plural than the data and facts permit. If it turns out that populations aren't the best unit of selection in explanations of clutch regulation in seabirds, then they aren't. We can't be a pluralist without some reason to think it is correct.

I have argued in print, rather opaquely, that species are also pluralities. I believe that differences in shared mechanisms and processes are what isolate species from each other, and generate cohesion within them. Sex won't isolate a parthenogenetic species from a sexual one. But needing a developmental process that at some point has a diploid egg is shared between two such species as sexual and asexual whiptail lizards. The difference is that one needs sperm to get that egg, and the other needs something else. That is what separates them.

But there are many ways of separating lineages. Some of them don't involve differences of the fertilisation system. Some of them involve mating behaviours, such as dances, calls, and search images. So we ought to define what makes a species in a given group of organisms in terms of the actual mechanisms that keep lineages distinct or cohesive. That is the "species concept" needed for that group. Each clade has a (historically derived) species "concept", or better, definition.

Now this raises an interesting question, which is most clearly brought out when we compare species from phylogenetically very distinct groups. Let's use an archaebacterial clonal species and a fully sexual eukaryote, as an example to base this question upon: are these two species commensurate? Can they be cross compared? In the asexual case, genome, organism, lineage and species seem to coincide (I think they don't, but assume they do). In the sexual case, they are all distinct. And we can locate many examples in between easily enough. Or, to put this another way, is an archaeal bacterial species worth the same as a sexual mammal species (a question that vexes conservation biology, for instance)?

Or take the concept of an "individual"? Debated since the late 19th century, the question of what counts as an individual is unresolved. Some opt for developmental cohestion, which sponges and slime molds don't have. Some opt for genetic identity of cell lineages, which stands of clonal trees, each of which can stand alone, do have. For any intuition, there's a counterexample waiting to be found (as Leo Buss showed in his book). As David Hull says, echoing Seneca on philosophers, nothing in biology is so weird that something, somewhere, doesn't have it.

All of these arguments, proposals and debate seem to me to rest on one simple premise - there is an objective, accurate, and theoretically useful universal set of ranks, levels or types in biology. And this seems to me to be one thing that we can resolve by the use of facts. There are no universals in biology, and everything shades into each other if you look hard and far enough. Not all heredity is genetic. Not all species are sexual or well-defined. Not all individuals are cohesive and form from a single fertilised fusion of haploid gametes.

Why is it that we expect to find such privileged single units of biology? I think there are two, closely related, reasons at least. One is that we always begin with the familiar. The prototype of a species, an individual, an heritable trait, is that which we commonly encounter, in ourselves and those organisms that matter most to us - pigs, goats, sheep, horses and cattle. Or which we see most frequently - birds, cats, dogs, bears. Or those which matter to us most economically - honeybees, grain species, trees. It is something of a default state of mind - as psychologist Susan Gelman noted in her The Essential Child, we are born essentialists and we expect the world to behave that way. Of course, this predisposition to infer that things have stable properties is a survival trait, and is reinforced by stability in the world about us as we grow, including the social world. But it is misleading when applied to broader biology. By "broader", I mean that when we encounter biology in contexts outside the limited range in which we grow up, and when we study it intensively rather than treat it as a simple background to our childhood lives, we find that things are not so simple.

In one way, biology is like geography. We need to identify peaks and troughs, active and static regions, of the living world. And a lot of the living world remains, for all intents and purposes, quite stable enough that our "maps" of it are useful. But as we expand beyond the personal experience of children and the village, our maps become more complicated and less easy to assign categories to, just as "mountain" has no objective meaning. In Australia, what we call mountains are small pathetic things compared to, say, the Rockies or the Andes, or the Himalayas, and they in turn are pathetic compared to the mighty mountains and canyons of Mars. Context and variation from small to large are important everywhere, and while we might not have any objections to calling these weathered remnants of ancient mountain building in Australia "mountains" (and some of them are only a few dozen metres high), it is hardly the case they are in the same league as Everest or Mount Fuji.

And that is the other reason why there are no fixed objective and non-arbitrary ranks or levels in biology. It needs to cover everything from the simplest virus to the largest superorganism and ecosystem, all the way to the biosphere. The need for fixed ranks is an outcome of our native essentialism. Which is not to say that we are "constructing" our world. If a mapmaker traces the peaks of sanddunes in the Negev desert, and they shift from year to year, it remains true that there are sanddunes there, and that they were in the location shown if properly mapped, even if we cannot identify the exact boundary between them. A Sorites heap does not necessarily cause the heap to be "constructed". Some things will be more or less exactly delineable; others won't. But there are no universal distinct and always operating categories for biological classes of things. So we should give up not the measurements of the things or the observation of them, but the expectation that they will fall into prior textbook-summarisable categories.

Some things are non-arbitrary. It is not arbitrary that a species splits in two, even if the exact moment, if there is one (and theory suggests there won't be), at which it happens is vague or uncertain. But at t, there is one coherent lineage, and at t + n, there are two, and that is not devalued because we know not the moment of the split or the future fate of those lineages. It is not arbitrary to define the transcription of genes, when that is clear cut. It is not arbitrary to identify an individual organism of, say, Homo sapiens, in terms of its developmental trajectory and relative coherence (although identifying and defining commensuals can get a bit tricky).

Units in biology are always group-relative, post hoc, and a posterori. They are phenomenal objects some of the time. They may in fact be undiagnosable or not measurable in any simple manner, but they should never be thought to exist solely in the mode that the prototype did, unless they are closely related to that prototype, and they haven't changed that mode yet. We have one genetic code, archaea have a slightly different one. Talk of "the" genetic code is unwarranted until we know for sure, or sure enough, what codes there are (presently numbered at 17 variants, I gather). Since biological knowledge is always defeasible by the very next thing one investigates, and no categories seem to be universally true of all life, real or possible, so too our categories should be the result of investigation rather than the buckets into which each case is forced.

So, abandon exclusivistic definitions of species, of ecosystem, community, gene, organism and organ. Use the terms, by all means, but understand that they apply to things only in virtue of their playing conceptual role, and not usually a theoretical role at that. In each case, there will be slight and subtle or large and obvious differences from the use of those categories elsewhere in biology, and that is a theoretical expectation. What we need our terms to deal with are the phenomena, and they come in a variety of kinds and levels.

To return to species for a minute, if species are evolved modalities of being distinct lineages, why call them all species? I would answer that there is no theoretical reason for this - a species is just a phenomenon in need of explanation, not a "level of organisation" that all and only species exhibit. A species is where genetic, organismic, haplotype and developmental lineages tend to coincide. When they do, we investigate the reasons that set of lineages coincides thus. The explanation is not necessarily that they form protected gene pools, though many will. Nor is it that they are ecologically and demographically replaceable sets of individuals, though they may be. Nor is it that they are able to recognise specific mates, though most will. Each case calls for a specific, if you'll excuse the pun, explanation, one that takes into account the unique historical processes for each species. If we can generalise these explanations over a group of species, all the better for us. In that case the generalisation will be due to either homology (plesiomorphies of shared ancestors) or homoplasy (convergently arrived at evolutionary traits). Homology-based explanations are generaliseable because of history, while homoplasious explanations rely upon some adaptive story. The problem with homoplasy is that all you can generalise is the trait being used to specify the class, while a homologous trait will be generaliseable with other traits of that clade.

In cladistics, or phylogenetic classification, what defines or identifies a monophyletic group (a group that includes an ancestral species and all of its descendants, no matter how modified they are) is a shared derived character. This is a rankless form of classification. A clade can include two or more taxa, and reflects only the branching order of whatever taxa are used in the classification (it may not "know" about many other unrecorded taxa in the evolutionary tree). Any number of taxa (which can be subspecific, although usually species or some surrogate for a species form the terminal nodes) can be a clade. Few have taken the next step, though, and argued that species is also a rankless taxon (one exception is Gareth Nelson, who argued this back in the 1980s). But why not? If a species is a phenomenon, and not a rank, then it explains why we do not think that a bacterial species and a mammalian species are commensurate. Commensurability of taxa, of species, is a relative property. If two taxa are closely related, then they are more easily compared to each other. If they are distantly related, or highly modified in what makes them each species, then they are harder to compare. As we should expect if species is not a rank or level, but merely what we see if we compare various kinds of lineages and note that they all tend to covary at that "level" for a given species.

Something like this is necessary to avoid dividing species into haplotype groups, simply because those groupings are monophyletic (and hence clades of genetic groups). We can divide our organisms quite a number of different ways, and to almost any level, right down to kin groups, that we like. Where to stop that is not arbitrary? That is why I think that species is a phenomenon, rather than a theoretical class concept or objectively specifiable level of organisation. Similar arguments can be given for other class concepts in biology, such as homology. If it is just a phenomenon that needs explanation in each case, then we no longer need to specify what level they are at - they can be a short sequence of DNA, a large karyotypic structure, cytological structures, tissue types, organs and skeletal or developmental structures and processes, and so on. They might even be population structures. Homology is on this view (proposed by Paul Griffiths) something we see on investigation, that we need to explain. It is not a particular kind of molecular or physiological or anatomical level of organisation. I am taking his idea and applying it generally.

We might want to cast this in terms of "holism versus reductionism", but I think this is somewhat different to that. To assert a holistic view, one needs to have a preset level that is the "whole", either for all organisms or some case under explanation. Instead I would suggest that this is about the data, and finding a phenomenal level that is interesting and calls for an account in terms of whatever happens to apply to that phenomenon. It does say that not everything reduces to genes, for instance, but it does not say that nothing reduces to genes. Instead, it allows for a multiplicity of perspectives, until we find one that satisfies our observations and theoretical apparatus. It takes Dawkins' and Williams' "genes-eye view" and allows for many such perspectives. It allows freedom of methodology and approach without being anarchistic. One is constrained by the observational and experimental data. Not everything goes, or at least, not all the time.

So I say, let's drop all ranks, privileged units, and universal classes in biology, and generalise only where we can, for only that is where we ought.