The Lagoon

by Armand Marie Leroi

  He launched his attack from all sides. In the Novum organum, 1620, he accused Aristotle of pressing his facts to fit his theories:

  Nor is much stress to be laid on his [Aristotle’s] frequent recourse to experiment in his books on animals, his problems, and other treatises; for he had already decided, without having properly consulted experience as the basis of his decisions and axioms, and after having so decided, he drags experiment along as a captive constrained to accommodate herself to his decisions: so that he is even more to be blamed than his modern followers (of the scholastic school) who have deserted her altogether.

  The Royal Society’s propagandists – Thomas Sprat (History of the Royal Society, 1667) and Joseph Glanvill (Plus Ultra, 1668) – echoed the charge. Glanvill was particularly caustic: ‘he [Aristotle] did not use and imploy Experiments for the erecting of his theories; but having arbitrarily pitched his Theories, his manner was to force Experience to suffragate, and yield countenance to his precarious Propositions’.

  Bacon’s most serious charge was aimed at Aristotle’s explanatory system. Of the four kinds of causal explanations that Aristotle insists natural science demands, Bacon ruled two – the formal and final – illegitimate. Natural philosophy should concern itself with the properties and movements of matter and them alone. Explanations such as ‘the hairs of the eyelids are for a quickset and fence about the sight’, or ‘the firmness of the skins and hides of living creatures is to defend them from the extremities of heat or cold’, or ‘the bones are for the columns or beams, whereupon the frames of the bodies of living creatures are built’ were no part of science, but should be left to metaphysics. They were ‘remoras and hindrances to stay and slug the ship from further sailing’. They retard the search for the true, physical causes of things.

  Bacon’s attack on forms was subtler. It is, he said, futile to inquire into the form of a lion or an oak or gold or even water or air. To the degree that forms have a place in natural philosophy, they are only a list of the basic sensible properties of matter, heavy–light, hot–cold, hard–soft and the like. His formal properties were grounded in a particulate (in seventeenth-century jargon, ‘corpuscularian’) theory of matter. Heat, for example, is a type of motion found when particles are both set in motion and constrained in some way. He evidently envisioned a heat ‘law’ that relates particulate motion to temperature. Bacon sidestepped the question of how to get more complex objects – gold or lions – out of these basic properties; but, then, he was more concerned with providing general principles than usable theories. The thrust, however, was clear: a radical, un-Aristotelian, ontological reductionism in which there is room only for moving and material explanations. Bacon looked for a new philosophical champion in antiquity. Democritus. He would become the Attic poster boy of the new scientific age.

  Bacon’s aversion to Aristotle and Aristotelianism – he scarcely distinguishes the two – also stemmed from a particular vision of the purpose of science and its proper object of study. Its purpose, in Bacon’s view, was not merely to understand the world, but to change it; its proper object of study, then, was the artificial rather than natural. Bacon was a technology enthusiast. Aristotle’s philosophy, he said, was ‘strong for disputations and contentions; but barren of the production of works for the benefit of man’. Bacon demanded a new, mechanistic natural philosophy underpinned by a unified physics that would explain the movements of both natural and artificial objects. Newton would provide one.

  In biology, the cheerleader of mechanism was Descartes. Animals and plants, he declared, do not have souls – they are merely machines. This was the doctrine of the běte machine or beast machine. Descartes reduced the complex of Aristotelian changes to local motion alone, and founded his physiology on a corpuscularianism that he got from Gassendi and Beeckman. His mathematical physics was important, but his anatomy indifferent and he made no biological discoveries. (He contested with Harvey over the movements of the heart and lost.) His teleology was simply theistic. (Animals may be machines, but they are wondrous machines made by God.) But his explicit comparison of animals to automata resonated at a time when mechanical devices were proliferating. It did away with the obscurities of the Aristotelian nutritive and sensitive souls (rendered utterly opaque by the schoolmen) and gave a point of entry for experimental investigation. In 1666 the Danish anatomist Niels Stensen (Steno) wrote:

  No one but [Descartes] has explained all human function, and above all those of the brain, in mechanical fashion. Others describe for us man himself; Descartes speaks only of a machine, which at one and the same time shows us the inadequacy of others and points out a method of investigating the function of the parts of the body just as insightfully as he describes the parts of his mechanical man [italics mine].

  The bête machine flexed its muscles and gave a lusty squall.

  Such then, in brief, are the intellectual currents that destroyed Aristotle’s science in the seventeenth century. His fortunes have varied since. Zoologists have always regarded him with affection. In the nineteenth century, Cuvier, Müller, Agassiz and many others even turned him into a bit of a cult.* To them, he was an illustrious forebear with a sharp eye for curious bits of zoology, an authority to wield against opponents, and even a fertile source of explanatory ideas – or so I have argued. In the eighteenth and nineteenth centuries, too, teleology was retrieved from the metaphysical and theological wastebasket to which Bacon and Descartes had consigned it. In some scientific circles, particularly German ones, final causes became respectable again. That was very Aristotelian. The consequences of that for his reputation were, however, ultimately malign. The association between teleology and vitalism revived and reinforced Bacon’s old charge that Aristotle’s science was unmechanistic. Hans Driesch, that errant embryologist, even wrote a history of vitalism that commenced proudly with Aristotle. Twentieth-century biologists would still be flogging vitalism long after it had expired. ‘And so to those of you who may be vitalists I would make this prophecy: what everyone believed yesterday, and you believe today, only cranks will believe tomorrow’ – Francis Crick in 1969. When, in 1954, Erwin Schrödinger published a little book on ancient science he simply stopped with Democritus. Why bother going further? Aristotle had nothing to say to modern science.

  CX

  BACON AND HIS successors said that Aristotle’s methods were wrong and that his explanations were too. Both charges are grave, but are they just? Our ideas of what constitutes scientific explanation, and how to achieve it, are ever changing. It may be, then, that we can see merits in Aristotle that our predecessors missed. Every generation must read Aristotle anew.*

  That Aristotle made countless observations of the natural world is obvious to anyone who reads his books – even the men of the Royal Society conceded so much. Should you read Aristotle’s biology, you may, however, wonder why Bacon and Glanvill keep going on about his ‘experiments’. They said he did them but abused their results, using them merely to confirm what he knew or thought he knew. You, however, may be less dismayed by his abuse of experimental data than by its absence.

  The difficulty is merely semantic. In the seventeenth century ‘experiment’ meant any investigation of a natural phenomenon that involved some sort of intervention. Aristotle’s study of chick embryogenesis which involved finding eggs of just the right stage, carefully cutting open the shell and prodding the embryo to expose its heart is, in this sense, one. Starving and then strangling livestock to see their vascular systems is another. So are vivisecting tortoises and poking out the eyes of swallows. Aristotle sometimes indicates that he’s actually tried it out by using the term pepeiramenoi: ‘Saltwater, when it turns into vapour, becomes sweet; and the vapour does not form saltwater when it condenses again. This we know by experiment.’ Or so pepeiramenoi is often translated.

  A modern scientist would take a more austere view. Such manipulations, he would say, are just observations made using a fancy technique. Experiments are defined not by their technique
but by their logical structure. A true experiment is the comparison of a deliberately manipulated situation to an unmanipulated control for the purpose of testing a causal hypothesis. And Aristotle’s works, he would sadly conclude, are devoid of experiments of that sort.*

  Why is this? Aristotle certainly understands experimental logic, for he repeatedly refers to what we would now call ‘natural experiments’. The oysters that were taken from Lesbos to Chios did not breed in their new home: ergo the generation of oysters depends not on the presence of oysters but on the right kind of mud: ergo they are spontaneous generators. The inference is plausible but far weaker than Aristotle allows. Perhaps Chian waters were too cold for the oysters to breed; perhaps they did breed, but the infant oysters died undetected; perhaps . . . a dozen alternative explanations spring to mind. Ecologists and evolutionary biologists often speak of ‘natural’ experiments since it’s hard to perturb the course of evolution or tweak whole ecosystems, but, as one of my colleagues, a famous ecologist himself, is fond of remarking, ‘The thing about “natural” experiments is that they’re not experiments at all’* – by which he means that in a true experiment the only variables that differ between control and treatment are those manipulated by the experimenter; and when you rely on nature to do your manipulations you can never be sure just what she’s meddled with.

  Theophrastus’ reports of how wheat cultivars perform when grown in various places are better and, done deliberately, would be a reciprocal common-garden experiment of the sort that would fully justify his inference that wheat strains differ due to some inherited quality. But he didn’t do them deliberately and so his inference, though very likely correct, is also weak. Who really knows what farmers get up to? If you believe them, you’ll end up believing all sorts of things; you’ll even believe that aira can evolve from wheat. Aristotle’s version of a common-garden experiment is even more pleasing: to determine whether the infertility of a couple is due to a deficiency in the male, he says, let him copulate with women other than his wife, and see if he sires offspring with them. Now that’s got the makings of a real experiment, and it would have been perfect had he also recommended, which he did not, the reciprocal treatment. But it’s no more than a suggestion. To imitate my colleague: ‘The thing about thought experiments is that . . .’

  It’s not as though the experiments were technically difficult. Do flies really spontaneously generate from rotten meat? All you need to test the idea are two jars, some fresh fish and a bit of fine cloth. That was the sum of Francesco Redi’s equipment. Does the embryo of a tetrapod truly emerge from a coagulum of semen and menstrual fluid? If so, then the coagulum should be visible in the dissected uterus of a freshly impregnated mammal; even a sheep would do. Aristotle didn’t look; William Harvey did.*

  Historians sometimes attribute Aristotle’s failure to do experiments to his worldview. If you draw, as Aristotle did, a sharp distinction between natural and unnatural change then a manipulative experiment, which clearly involves the latter, can hardly shed light on the former. There may be something to this. In the centuries after Aristotle’s death Greek technologists in Alexandria began to produce elaborate machines. In the first century AD Hero of Alexandria described a charming hydraulic device in which a cluster of bronze birds cease to sing as a bronze owl rotates to face them. The gadget-minded Alexandrians were also quicker than Aristotle to put their physical theories to the test. Hero’s Pneumatics contains an account of an experimental programme that is almost worthy of Boyle.*

  Perhaps, then, the reason why Aristotle did not roll a ball down an inclined plane as Galileo did is because the conceptual structure of his physics prevented him from doing so.* But does that explain why he did not ask a farmer to mate a white-fleeced ram to a black-fleeced ewe to see how the progeny turn out? It’s not a particularly ‘forced’ intervention, he understood the logic (vide his discussion of that wayward woman of Elis) and the results of the experiment would surely have given him pause for thought when constructing his model of inheritance.*

  Indeed, had Aristotle done a few simple experiments, he would certainly have made fewer errors. But it’s one thing to understand experimental logic, quite another to see it as the high road to truth. The question, however, is this: given that he didn’t do experiments, does his method correspond to anything that we, today, might recognize as science? Plato’s method plainly disqualifies his theories from being scientific ones – it was, after all, founded on a contempt for empirical reality. It’s harder to be certain about the physiologoi’s methods – they are so diverse and we have so little idea what, exactly, they did. Aristotle, however, has a method for extracting truth from the empirical world, a very sophisticated one too. It is, I believe, very similar to one used today.

  CXI

  SCIENCE HAS ALWAYS embraced two very different styles of empirical investigation. The first is the style most familiar to us, in which causal hypotheses are tested by deliberate, critical experiments. It’s the style adopted, and celebrated by, the founders of the Royal Society. The second is less familiar, but hardly less important. It’s one in which data are amassed, patterns sought and causal explanations inferred from those patterns. It is the style that was once found only in the historical sciences – cosmology, geology, palaeontology, ecology and evolutionary biology – the sciences in which manipulative experiments are hard to do. That, however, is no longer true.

  The first style dominated twentieth-century biology. First you identified the object of your study – some gene (say) in some creature that, for whatever reason, you thought particularly fascinating. You worked out a way to measure its activity. Then you manipulated it. You could ‘knock out’ your gene – kill it dead in its tracks – or ‘over-express’ it – turn it on in unexpected ways and places. You would see how your manipulation affected the gross phenotype of your creature or, perhaps, the behaviour of other genes – but not too many of them since each test was complicated, expensive and time consuming. All this would take years. When you were done you’d publish a paper like this one:

  Morita, K. et al. 2002. A Caenorhabditis elegans TGF-beta, DBL-1, controls the expression of LON-1, a PR-related protein, that regulates polyploidization and body length. Embo J. 21:1063–73

  The authors of this paper – and there are thousands like it – compared long mutant worms with regular worms and so described the role of a few genes in controlling the length of their worm. They know full well that they’ve unravelled only a few links in a vast causal network that might influence worm length, but since they believe in their results, they rest secure in the knowledge that their causal claim is true; and that, modest though their discovery may be, when it is combined with a thousand others like it, something important will emerge.

  That paper was published just over a decade ago. How dated it now looks. For, as the twenty-first century has progressed, the notion of studying just a few genes at a time has become quite passé. The problem is no longer how to find a gene of interest – a single, modern, genome-sequencing machine can pump out fifty-four gigabases of sequence per day.* That’s around sixteen human genomes, each with 25,000-odd genes. An ‘expression array’ chip can show, for any tissue you can grind up and put on it, which of those 25,000 are active and to what degree. Other technologies will allow you to survey thousands of metabolites or proteins all at once. Biologists speak of the ‘omics – genomics, transcriptomics, metabolomics, proteomics, but they all really mean just one thing: data – lots of it.

  Here is a typical ‘omics paper:

  Fuchs, S. et al. 2010. A metabolic signature of long life in Caenorhabditis elegans. BMC Biology 8:2

  The authors of this paper – and there are thousands like it – compared long-lived mutant worms with regular worms and described many differences in their metabolites. This paper (Fuchs) has much in common with the one above (Morita): same worm (C. elegans); same lab (mine); similar problems (growth v. ageing). But there’s a fundamental difference in method. Where Morita s
tudied a few genes in detail, Fuchs studied hundreds of metabolites superficially. The consequences of this for what they could claim were profound – and were reflected in the titles. Where Morita spoke boldly of ‘control’, Fuchs admitted only to having found a ‘signature’. She found dozens of differences between their long-lived and short-lived worms, but had no idea which of them matter, which – if any of them – actually make the long-lived worms live long. Technology gave Fuchs comprehensiveness. It cost her causality. Of course Fuchs and her colleagues did not despair. They – we – trawled her data for patterns. We found them and, from them, wove a causal model that we fondly believe might contain some truth. But we’d be the first to admit that we really don’t know.

  Fuchs’ paper is very much in the second style. It’s the characteristic style of the age of Big Data, one that is spreading into sociology, cultural history, engineering and economics. Its method is always the same: hoover up all the data going, order it by some sort of classification, visualize its structure, infer a causal model. The tools – multidimensional scaling, network graphs, Self Organizing Maps and the like – are all new, but the style is old.

  It’s Aristotle’s style. It’s the style to adopt when you discover a new world; when, instead of worrying away at the phenomena your predecessors looked at, you chance across some vast new domain of things to study, splendid in their confusion, fascinating in their order and opaque in their causes. These days technology – better sequencing machines, faster computers, bigger telescopes – give us our new worlds. Aristotle needed none of them. He just had to walk down to the shore to find a whole domain of things that had never been studied before. Historia animalium is the Big Data repository of his day. Not so big, you say? Maybe – but his other Big Data project, the 158 constitutions he collected, would be impressive even now. They were the basis for his other great exercise in causal explanation, the Politics.