8

Experiments

Thus the discovery of the barometer transformed physics, just as the discovery of the telescope transformed astronomy … The history of science has its own revolutions, just like the history of nations … with this significant difference, that revolutions in science … successfully achieve what they set out to do.

– Vincenzo Antinori, ‘Notizie istoriche’ (1841)¹

§ 1

On 19 September 1648 Florin Périer, brother-in-law of the French mathematician Blaise Pascal, accompanied by a group of local dignitaries from Clermont-Ferrand, set out to climb the Puy-de-Dôme in the Massif Central.ii² Below them, in a monastery garden, they had left an inverted tube sitting in a bowl of mercury. The height of the mercury in the tube was just over 26 inches (they measured in pouces, or inches, but their inches were very slightly longer than English inches). When they reached the summit, some 3,000 feet higher by their calculation, they mounted another barometer (as we would call the instrument: the word, in both English and French, dates to 1666, preceded in English a year earlier by ‘baroscope’). The height of the mercury in the tube was a little more than 3 inches lower at the summit than in the garden of the monastery, and they obtained the same result when they took their barometer apart and reassembled it in several different places on the summit. On the way down they repeated the experiment a couple of times at a point nearer the foot than the summit of the mountain: there the mercury was an inch lower than in the monastery garden. One of these repetitions was carried out by a M. Mosnier. The next day they carried out the same experiment at the base and at the top of the tower of the cathedral church in Clermont: the difference was small (about two tenths of an inch) but measurable. Pascal, learning of this last result, carried out similar experiments in tall buildings in Paris and, as quickly as possible, published an account of them. Looking back in 1662, Boyle hailed the experiment at the Puy-de-Dôme as the experimentum crucis, the crucial experiment, which had validated a new physics.³ Indeed, this was the first experiment to be hailed with this phrase, later made famous by Newton when he used it in connection with his prism experiments, which demonstrated that a ray of white light is made up of a spectrum of coloured rays.⁴

This is the first ‘proper’ experiment, in that it involves a carefully designed procedure, verification (the onlookers are there to ensure this really is a reliable account), repetition and independent replication, followed rapidly by dissemination.⁵ The experiment was intended to answer a question: was there some natural resistance to the creation of an apparently empty space at the top of the tube (because, as Aristotle claimed, nature abhorred a vacuum), or was the height of the mercury (and so the size of the empty space) determined solely by the weight of the air? Pascal always claimed to be the inventor of this experiment, but the philosopher René Descartes insisted that he had originally suggested it to Pascal, and their joint friend Marin Mersenne was busy trying to organize the very same experiment when Pascal beat him to it. (Mersenne had trouble getting hold of sufficiently long, robust glass tubes ‘hermetically sealed’ at one end, although he seems to have gone to the same supplier as Pascal, who had no difficulty – perhaps Pascal was buying all the tubes that could be made.)iiii

There was general agreement that the experiment showed that the height of the mercury was determined by the weight of the air, but there was no agreement as to whether the space at the top of the tube was properly considered to be a vacuum: Pascal thought it was, but Descartes thought it contained a weightless aether (without which, he thought, light would be unable to cross from one side of the tube to the other) capable of passing through glass; and Pascal’s friends Mersenne and Roberval thought that there was some rarefied air in the space. This story is normally told as though Pascal was right and Mersenne and Roberval wrong, although actually all three were right: the space was in effect a vacuum, but it did contain some air under extremely low pressure.⁶ Pascal’s interpretation of the experiment was in direct contradiction to Aristotle’s claim that nature abhors a vacuum.

If we have something they didn’t, then experiment would seem a good candidate. As we saw in the last chapter, it is not always easy to define the threshold for saying that a culture ‘has’ something, but language usually provides a useful marker. This is less true when we turn to experiments. Experientia and experimentum (‘experience’ and ‘experiment’) are more or less synonymous in classical, medieval and early modern Latin, and all the modern languages that have both words initially reflect the Latin usage.⁷ In modern English the distinction is clear: going to the ballet is an experience; the Large Hadron Collider is an experiment. But this distinction emerged slowly and became firmly established only during the course of the eighteenth century. The OED gives 1727 as the last date at which ‘experiment’ as a verb was used to mean ‘experience’, and 1763 as the last date at which ‘experience’ as a noun was used to mean ‘experiment’.iiiiii Insensitive to this change of meaning, scholars frequently translate the word experimentum in Latin texts as ‘experiment’, thus often giving a totally false impression of its meaning, which is commonly ‘experience’.

Something close to the modern distinction is, however, to be found in Francis Bacon, who distinguishes between two sorts of experience: knowledge acquired randomly (by ‘chance’) and knowledge acquired deliberately (by ‘experiment’).iviv By this definition, though, going to the ballet is an experiment, while learning that the seats are uncomfortable and the drinks sold in the bar overpriced is a chance experience. Moreover, it is quite wrong to think that Bacon is an advocate of an experimental (in our sense), as opposed to an experiential, science. He does think experiments can supplement experiences and provide crucial information, but he attacks William Gilbert for studying the magnet through a narrow experimental programme which concerns itself only with magnets: ‘For no one successfully investigates the nature of a thing in the thing itself; the enquiry must be enlarged so as to become more general.’⁸ Hobbes, for his part, neatly distinguishes experiment from experience, but not as we do. For him, several experiments amount to experience – experiment is particular; experience general.⁹

At first sight one would think that Henry Power’s Experimental Philosophy of 1664 is a book about experiments in the modern sense, and indeed it includes numerous experiments involving mercury and glass tubes; but the first section of the book is concerned with ‘experiments’ made with a microscope. Power is, however, well on the way to our modern usage, because, although he says the book is about ‘new experiments, microscopical, mercurial, magnetical’, he labels each section of his microscopical reports an ‘Observation’, and each section of his mercurial reports an ‘Experiment’. ‘Observation’ used in this modern sense (rather than in the sense of a practice, such as a religious observation or observance) was relatively new in English, although it exists in classical Latin (observatio): the OED gives 1547 as the first usage of ‘observation’ and 1559 as the first of ‘observe’ in this new sense. Over time, observation became an adjunct to experiment, both producing reliable facts in place of the unreliable, unspecific ‘experience’ which underlay so much classical and medieval discussion.¹⁰

In French and Portuguese the old confusions (as they must seem to an anglophone) still exist. In French there is a verb, expérimenter, which corresponds to both ‘to experience’ and ‘to experiment’; there is still no noun that corresponds to the English ‘experiment’, although you can faire une expérience, where expérience means ‘experiment’, and in nineteenth-century French expériences in the plural always meant experiments, not experiences. French has also acquired the word expérimentation, which is now sometimes used as if it were equivalent to ‘experiment’.vv¹¹ There is also in French (and Portuguese) an adjective, expérimental, classically used in the phrase philosophie expérimentale. The word expérimental was used solely in a religious, usually a mystical, context until the translation of Sprat’s History of the Royal-Society into French in 1669, when the phrase philosophie expérimentale was introduced into the language. It is something of a puzzle that the word ‘experiment’ did not come with it, despite its respectable Latin antecedents.

There are other words just as ambiguous as sixteenth-century ‘experience’/‘experiment’. A striking example is ‘demonstration’. In classical Latin you demonstrate something by pointing it out with your finger. But in the Middle Ages the word demonstratio was used to refer to a deduction or proof in philosophy or mathematics: thus you can demonstrate or prove that all the angles of a triangle add up to two right angles. In French this remained the meaning of the word until very late: it is only in the fourth edition of the Dictionary of the Académie Française (1762) that the use of the word in contexts where you show someone what you are talking about (a demonstration in anatomy, for example) is recorded. In English the two meanings (‘demonstration’ as deduction; ‘demonstration’ as pointing out) exist together from an early date. Thus both Aristotelian philosophers and the new scientists produced demonstrations, but they meant radically different things by the word.

Another striking example is ‘proof’. On the one hand, we use the word to refer to proofs, deductions and demonstrations in mathematics, geometry and logic. On the other, we talk about ‘the proof of the pudding’, alcohol being 40 proof and proving a gun. ‘Proof’ thus covers both necessary truths and practical tests, and it has the same etymological root as ‘probe’ and ‘probability’. This ambiguity comes from the Latin (probo, probatio) and is found in all the modern languages derived from Latin (Spanish probar, Italian provare, German probieren, French prouver – though in French there is also éprouver, to test, so in modern French prouver has lost the sense of ‘to test’). A proof, at least in mathematics and logic, is an absolute; you either prove something or you do not. On the other hand, evidence (to use our modern English word) is something you can have more of, or less of. In Roman law two witnesses may provide a full proof of guilt; one witness and a confession may do the job; or one witness and circumstantial evidence (the accused’s knife, for example, was found in the victim). Renaissance lawyers were trained to talk about half a proof or a whole proof.

In cases where a proof was not complete and there was no alternative way of obtaining evidence, torture was (in countries which followed the principles of Roman law) legally used, from the thirteenth century to the eighteenth, in the hope of acquiring complete evidence. In the case of della Porta, for example, the tribunal of the Inquisition voted to torture him mildly (leviter) in view of his poor health; then a week later, luckily for della Porta, they changed their minds. We have no record of della Porta’s thoughts and feelings during this intervening week.¹² Perhaps he became so ill at the prospect of torture that he could no longer be tortured (for you had to pass a medical examination and be declared fit before you underwent torture; the Inquisition was scrupulous about such matters). Someone against whom an incomplete proof of guilt existed (someone who had been tortured without confessing – Machiavelli in 1513, for example) was neither guilty nor innocent but could properly be punished for having given grounds for suspicion (this is what happened to della Porta, and to Galileo when he was tried by the Inquisition in 1633; Machiavelli had the good luck to be released under an amnesty). Francis Bacon, when he writes about experiments, uses the phrases ‘the inquisition of nature’ and ‘nature vexed’. Does this mean torturing nature to extract an answer?¹³ Bacon had himself seen men suspected of treason racked, though torture was not normally used in English legal proceedings. In a world where legal metaphors were constantly employed when knowledge was being discussed (as we have seen, the word ‘fact’ is itself a dead legal metaphor) questions of proof always carried with them the possibility of torture as a (metaphorical) mode of proceeding, but in English law ‘inquisition’ (an inquest is an inquisition) and ‘vexation’ carry no necessary implication of torture.

William Gilbert, writing On the Magnet in Latin, is, as one would expect, alert to the difficulty of using words like ‘proof’ and ‘demonstration’ to describe what experiments do. His preferred term is the post-classical word ostensio, a display or showing, which he defines as ‘a manifest demonstration by means of a body’. In other words, he is not providing a demonstration in the logical or mathematical sense, but he is making some physical reality apparent. He intends, he says, to show you things as if he were pointing to them with his finger. When you read his book you are a ‘virtual witness’ of his experiments.¹⁴

§ 2

This chapter began in 1648, with Pascal’s Puy-de-Dôme experiment, but Pascal was not the first experimental scientist. Take, for example, the evolution of Galileo’s thinking on the question of buoyancy. He started out as an admirer of Archimedes. In an early unpublished text from the 1590s he sought to demonstrate that Archimedes’ principle, that a body floats when it displaces its own weight in water, is necessarily true.¹⁵ The text of Archimedes had been available in Latin from the twelfth century, and had been first published in 1544. Early editions of Archimedes come with illustrations which show objects floating in a vast ocean of water, an ocean which stretches right around the globe, and Galileo drew such sketches in his own text.

It is perfectly correct to claim that in a boundless fluid a floating body displaces its own weight in water. But as he revised his text Galileo went on to represent objects floating in containers, such as a tank standing on a table top. When you put a block of wood into a tank, the level of water in the tank rises. Galileo at first thought that the volume of water above the old surface corresponded to the volume of water displaced by the object, and the weight of the water above the old surface corresponded to the total weight of the object, according to Archimedes’ principle. As we shall see, this is false. Unlike previous interpreters of Archimedes, Galileo had asked himself what sort of experimental apparatus would serve to illustrate Archimedes’ principle; what he hadn’t grasped was that this apparatus would serve to show that Archimedes’ principle is incomplete.

Twenty years later, in 1612, Galileo found himself in a dispute with Aristotelian philosophers. Heavier-than-water objects, they assured him, float if they have the right shape. Thus a chip of ebony, which is heavier than water, floats if placed on the surface of a bucket of water. Provoked, Galileo embarked on a series of experiments to study floating bodies. Chips of ebony, he found, float if they are dry to start with and are placed gently on the surface of the water – but so do metal needles. If they are already wet all over, they sink. Galileo was exploring the phenomenon we call surface tension.

Galileo also wanted to construct an object which would fully immerse but not sink – an object with the same specific gravity as water. He took some wax, mixed it with iron filings and shaped it into a ball: when he had the mixture right it floated just below the surface of the water. In this case, he wrote, in a draft text, the volume and weight of the water displaced correspond to the volume and weight of the ball, according to Archimedes’ principle – except they don’t. He was still repeating his old mistake.

At this point, Galileo had a sense that something was wrong. He went back to his old thought experiment and began to study it carefully, this time with the help of real tanks, real blocks of wood and pieces of marble. He tried floating the same block of wood and sinking the same block of marble in three different tanks, and worked out the mathematical formula which determines the extent to which the water in the tank is raised by the introduction of the blocks. He now understood the issue of displacement in terms of volume, and it was an easy step then to grasp it in terms of weight.¹⁶ It was now clear to him that when a block of marble is put into a tank so that only part of it is submerged it does not displace water equivalent to the volume of that part of the block which ends up under water; it displaces water equivalent only to the volume of that part of the block which is under the original surface level. Consequently, a block of wood floating in a tank displaces less than its own weight in water. According to Archimedes’ principle, if water were to occupy the volume of the block which is under water, that water would weigh the same as the whole block. Archimedes’ principle did not apply.

Galileo proceeded to confirm his new theory with a very simple experiment.¹⁷ He took a small rectangular tank and put into it a large block of wood which fitted snugly. He poured in water until he reached the exact moment when the wood began to float. He was trying to show – and succeeded in showing – that the ratio of the depth of the water to the overall height of the block corresponds to the ratio of the weights of equal volumes of wood and water. But he was also showing something very odd, which followed from his new discovery: a very small amount of water could be made to float a very large and heavy object; indeed, the actual water in the tank could weigh less than the block of wood it was lifting – which, according to Archimedes’ principle as traditionally understood, was impossible. (You can do this yourself by putting a small amount of water inside a wine cooler and then floating a bottle of wine in it.)

Now, at last, Galileo had a secure grasp of the principle that when you introduce the block of wood into the tank and the level of water in the tank rises, the volume of water displaced corresponds only to the portion of the block below the old, lower water-line, which is much less than the space occupied by the portion of the block which is below the new, higher water-line. The more closely the tank fits around the block the more powerful this effect, because the water is not being displaced sideways by the introduction of the block (as it would be in a boundless fluid), it is being raised upwards. Galileo had established that the relationship between the weight of a floating object and the weight of the water displaced by it in a bounded container is not comparable to that of two weights at either end of a balance but rather to that of two weights at either end of a lever. Archimedes’ principle is a limit case, not a universal principle. Without intending to, Galileo had devised an elementary hydraulic press.vivi

Galileo published these results in 1612, and they provoked a brief flurry of debate, but they were unnoticed outside northern Italy. The philosophers weren’t convinced and continued as before, and the mathematicians weren’t impressed – this wasn’t maths as they understood it. Galileo had been an experimental scientist for some time, a decade or so – in fact, since he had read William Gilbert’s On the Magnet. But this was the first time he had published the results of a series of experiments. Gilbert and Galileo were developing a new type of science, based upon systematic experimentation. But very few people paid any attention.

§ 3

There is nothing new about the idea of testing a theory; it is perfectly easy to show that Ptolemy and Galen had carried out experiments, and the Optics of the first great experimental scientist, Ibn al-Haytham (965–c.1040), had been translated into Latin by 1230 (at which point Ibn al-Haytham acquired his Western name of Alhazen).viivii¹⁸ It was soon widely available in manuscript and appeared in print in 1572. The puzzle is why Ibn al-Haytham’s example was not followed more extensively, for it would be difficult to overestimate the significance of his achievements. Using a rigorously experimental method, he refuted the standard extromission theory of sight (that sight is made possible by rays that go out from the eye) and defended the intromission theory (that sight is made possible by rays that enter the eye from the object); he produced the first full statement of the law of reflection, and also studied refraction; he designed the first true camera obscura; he made enormous advances towards an understanding of the physiology of the eye (although he failed to grasp that an upside-down image is projected through the lens on to the retina at the back of the eye); and he laid the intellectual foundations of the science of artificial perspective. Medieval optics was heavily dependent upon his contribution and he was unquestionably the best example of an experimental scientist before Gilbert.viiiviii

If Ibn al-Haytham offered plenty of real experiments, medieval philosophy is also full of thought experiments designed to test the implications of theories.¹⁹ What would happen, for example, if you drilled a tunnel down through the centre of the Earth and then dropped an object into the tunnel? Would it stop when it reached the centre, its natural resting place? Or would it shoot past? Would it oscillate back and forth until it came to rest? Obviously, this thought experiment is not one that could be carried out in practice (and no one tried to use a pendulum as a substitute),ixix but often experiments are described in a way that makes it difficult to tell whether they have actually been performed or not, and this continued to be true in the seventeenth century. Boyle complained that Pascal described experiments (carried out under 20 feet of water) that he could not possibly have performed, and modern historians have made the same complaint against Galileo (although, it has to be said, nearly always mistakenly).²⁰

The puzzle is therefore not whether there was any experimental science before the Scientific Revolution, for it is easy to find examples; rather, it is why there was so little, particularly given the example provided by Ibn al-Haytham and the prevalence of thought experiments. It is not difficult to identify a number of relevant factors.

First, experimentation involves manual labour. Although it has been argued that Christianity, particularly the monastic tradition, placed a higher value on it than had the ancient world, there was still considerable resistance within medieval and indeed Renaissance culture to physical work. The first experimenters were happy to use their hands. We are told that Galileo loved making little machines as a child (Newton certainly did),²¹ and that Torricelli was very skilled with his hands. Experimentation was a hands-on business.

Second, the dominant position acquired by Aristotelian natural philosophy within the medieval universities resulted in a double inhibition on experimentation. First, where Aristotle had discussed a subject at any length, it was assumed that adequate knowledge of it was already available (one reason why optics could develop as an intellectual discipline was that the first important treatment of the subject was by Euclid, not Aristotle); and second, the Aristotelian tradition insisted that the highest form of knowledge was deductive, or syllogistic, knowledge.

Medieval philosophers such as Robert Grosseteste (c.1175–1253) worked out a fairly sophisticated account of how one might work from experience to theoretical generalization, and then use theoretical generalizations to deduce the facts (or, rather, phenomena) of experience. But the whole point was that this procedure was only to be employed when there were no obvious first principles from which to work, and that it was (perfectly correctly) seen as being entirely compatible with Aristotle’s understanding of scientific knowledge. Thus Grosseteste claimed that we can know that all movement in the heavens is circular from first principles (if the movement was not circular, empty space would be opened up between the heavenly orbs, and this is impossible, as a vacuum is impossible), but we cannot deduce the shape of the Earth from first principles. Consequently, we must fill this gap by relying on experience, and experience provides convincing evidence (for example, eclipses occur earlier in the day at points further to the east and later in the day at points further to the west; the Pole Star sinks towards the horizon as one travels south) that the earth is spherical.²²

Experience and experiment are thus to be invoked only to fill gaps in a fundamentally deductive system of knowledge, never to question the reliability of deductive knowledge itself; and these gaps were always of limited significance within a curriculum centred upon Aristotle’s texts. (Grosseteste’s view that the shape of the earth was a purely empirical question was not without consequences, for it opened up the intellectual space for him to adopt a one-sphere theory of the Earth.) Grosseteste’s own practice demonstrates a remarkable indifference to experimental procedure; thus he formulated a general principle of refraction, but he simply assumed that it must, like the law of reflection, involve equal angles, and never conducted the elementary tests which would have shown that this assumption was misplaced. He produced a new theory of the rainbow which emphasized the role of refraction, where Aristotle had only mentioned reflection; but there is no evidence that Grosseteste ever conducted experiments to test his theory.²³ In 1953 Alistair Crombie published a book entitled Robert Grosseteste and the Origins of Experimental Science. Over the course of his life Crombie slowly retreated from the claims made in that book. By 1994 he was prepared to write:

It is difficult to tell whether so independent a thinker as Robert Grosseteste saw himself as doing and finding out something novel, beyond his authorities[,] as distinct from discovering their real meaning. It seems unlikely. Roger Bacon [1214–94, a follower of Grosseteste, and often heralded as an exponent of experimental science in the Middle Ages] saw contemporary scientific work as a recovery of ancient forgotten knowledge. Perhaps from this mentality, as well as from uncritical literal copying, came the medieval habit of reporting reported observations and experiments as if they were original.²⁴

Third, experimentation involves both a study of the external world and a capacity to generalize. It requires an ability to move back and forth between the concrete and the abstract, the immediate example and a scientific theory, and this movement is conceptually and historically problematic. The Greeks never thought of knowledge (episteme) as being knowledge of the external world, because for them reason was always universal and eternal; the mind was as one with what it knew.²⁵ In the Middle Ages, for example, Grosseteste adopted a neo-Platonist view that true knowledge was based on illumination, and the perfect form of knowledge was that of the angels, who needed no sensory experience of reality to know the divine mind and the universe through it.²⁶ This had a continuing influence into the early modern period: Descartes tried to recapture a Platonic conception of knowledge as being that which is self-evidently true, and even Galileo sought whenever possible to present his new sciences as mathematical demonstrations, not empirical extrapolations. Within this tradition knowledge is primarily mental, conceptual, theoretical and, in the end, mathematical.

Mathematicians, as members of an intellectual discipline, were thus torn between two types of knowledge: Plato and Euclid appeared to justify a purely abstract, theoretical form of knowledge; while the applied sciences of astronomy, cartography and fortification encouraged an empirical, practical orientation. Among the ancients, Archimedes appeared to have bridged this divide by showing how theory could be put to practical purposes, but the tension between the two approaches still continued in Newton, who sought to present his knowledge, in so far as possible, as pure theory, while insisting that it was based on evidence and had practical applications.

Catholic Christianity, in contrast, was committed to the belief that the truth lies outside us: the crucifixion of Christ and the transubstantiation of the host during Mass are not events in the mind but in the external world. Aristotle was thus interpreted by philosophers as grounding knowledge in sensation, and sensation was reinterpreted as knowledge of a reality external to the perceiver. But (and it is a big but) the truths of religion are not normally amenable to sensory perception; in the Mass, the bread and the wine continue to look like bread and wine. Hence the importance of miracles, when sensory perception confirms divine truth.

The medieval emphasis on external reality opened the way to nominalism, which, in reaction against Platonism and Platonizing interpretations of Aristotle, insisted that only concrete individuals exist and that abstractions are merely mental fictions – but in doing so it left little scope for a movement back from the particular to the general. Things are as they are not because of any sort of natural order or necessity, but because God chose to make them so. The world itself is a sort of miracle, and what happened yesterday need not happen tomorrow.²⁷

Experimentation thus required a deeply problematic balancing act between Platonic idealism and a crude empiricism. Experimenters have to insist on the particularity of experience, but they also have to claim that general conclusions can be drawn from specific examples. Underlying experimentation, therefore, there must be a theory of the regularity and economy of nature; the natural world has to be the sort of world that could, in principle, be interpreted through experiments. ‘For who doubts,’ asked Newton’s associate Roger Cotes, ‘if gravity be the cause of the descent of a stone in Europe, but that it is also the cause of the same descent in America?’²⁸ In addition, we have to be equipped to interpret the world; our senses have to pick the features which matter: Diderot doubted whether a blind person could ever come to recognize the universe as orderly, and so as divinely created. When the experimental method is successful in explaining what was previously inexplicable, it thus not only establishes particular scientific theories, it also confirms the validity of the general approach which underpins experimentation. Successful experimentation builds confidence in the experimental method; failure undermines it.

A further problem was that an experiment is an artefact. Aristotelian philosophy drew a sharp distinction between the natural and the artificial: understanding one provided no basis for understanding the other. In some cases this is obvious: a kite won’t help me understand how a bird flies, or a steam engine how a muscle works. Natural objects have, for an Aristotelian, their own internal formative principles, while artificial objects are made according to a design imposed from outside. The distinction between nature and artifice went even further: it was assumed that the rules governing the behaviour of artificial objects were different to those that operated in the world of nature, so that a machine might enable one to cheat nature by getting out more work than one put in. Galileo was the first to show that this could never happen.

It is evident that we can often understand what we make better than we can understand what nature produces, and this principle can be extended, for example, to mathematics, where we determine the rules of the enterprise. Thus in 1578 Paolo Sarpi wrote:

We know for certain both the existence and the cause of those things which we understand fully how to make; of those things which we know by experience alone we know the existence but not the cause. Conjecturing it then we look only for one that is possible, but among many causes which we find possible we cannot be certain which is the true one.²⁹

Sarpi gives mathematics and clocks as examples of knowledge where we have certainty because we have made what we know, and astronomy as an example of knowledge where we can come up with a possible right answer (the Copernican system, say) but can never be sure that it is correct. Sarpi never shared his friend Galileo’s conviction that Copernicanism was obviously right.

Such an approach implies that the sort of knowledge obtained by experiment need not be a reliable guide to how nature works. The fact that I can make a vacuum in the laboratory, for example, need not mean that a vacuum can ever occur in nature. It is often said that William Harvey demonstrated that the heart is a pump; but in De motu cordis he never compared the heart to a pump – pumps are artificial, after all, and hearts are natural. It would be dangerous to rely on such a comparison.³⁰ In contrast, the ‘maker’s knowledge’ principle implies that if I do make a vacuum in a laboratory then I have a real understanding of what it is that I have made.³¹ Confidence in experimental knowledge thus requires the natural/artificial distinction to be undermined and replaced by the conviction that by performing procedures that correspond to natural processes I can have true knowledge of those processes.

The first person to insist as a matter of principle that knowledge of artefacts could count as knowledge of nature was Francis Bacon, who said ‘artificial things differ from natural things not in form or essence, but only in the efficient.’³² Thus, knowledge of an artificial rainbow gives you (as we shall see in a moment) causal understanding of a natural rainbow, even though you have produced the artificial rainbow by different means. In a case like this the experimental method requires you to move smoothly back and forth between nature and artifice. Gilbert claimed that the little spherical magnets that he used were equivalent to the Earth; Pierre Guiffart, who had observed Pascal’s first vacuum experiments, said of the Torricellian tubes, ‘In them one sees a little miniature of the world,’ in that one actually sees the weight of the air.³³ Such claims were not straightforward – Jesuit scientists strenuously opposed Gilbert’s claim that the Earth itself was a magnet, and opponents of the vacuum protested that the Torricellian tube was deceptive, for it appeared to contain nothing in the space above the mercury when it surely contained something.

To a certain degree, if the world is orderly and predictable, it is because we have worked to make it so by developing technologies that give us control over nature. If we can model its processes, it is because we have developed our own capacities for making nature-like artefacts. It was therefore inevitable that the advocates of the experimental method in the seventeenth century would come to insist that the universe is like a clock, for clocks are embodiments of the principles of order, regularity and efficiency and, moreover, it is we who have made them. If we think of God as a clockmaker, then we can be confident that he will have made a world amenable to experimental enquiry. In the Middle Ages the heavens had been compared to clockwork; now the same principle of regularity was, it was claimed, to be discovered in the sublunary world.³⁴

Finally, there was, of course, in the Middle Ages, no culture of discovery. Even Ibn al-Haytham’s discoveries were hard to integrate into a system of knowledge that was backward looking, so the extromission theory of vision continued to be the standard theory simply because it was the one upheld by the authors who had antiquity on their side.

These five factors help explain the limited success of experimental science in a medieval context. Take, for example, Theodoric of Freiberg (c.1250–c.1310), who carried out the most remarkable experimental work in the whole of the Christian Middle Ages. Theodoric provided the first satisfactory account of the rainbow.³⁵ This involved direct criticism of Aristotle.xx Aristotle had said that rainbows are the result of reflection, while Theodoric showed that they were the result of two refractions and two reflections within each drop of water. Aristotle had denied that the colour yellow is really present in the rainbow, and had identified only three colours; Theodoric insisted that yellow was a fourth colour in the rainbow. Theodoric’s analysis depended partly on examining rainbow-like images that he encountered in daily life: in the spray thrown up by a turning water-wheel, in the drops of dew on a spider’s web. But he also studied what happened when a ray of light entered a glass ball filled with water, on the theory that this would provide a good model of what happened when a ray of light entered a raindrop (he used a urine bottle, which was a standard piece of equipment for any medieval doctor and had a spherical bulb). Around the same time a similar experiment was independently performed within a camera obscura by Kamal al-Din al-Farisi, who was, like Theodoric, drawing on the example presented by Ibn al-Haytham, and who will also have had access to urine bottles.³⁶

The illustration accompanying the late-thirteenth-century Theodoric of Freiberg’s study of the rainbow when it appeared in print in Trutfetter’s textbook of 1514. It shows that in forming a rainbow each ray from the sun is twice refracted and twice reflected as it passes through a drop of water before reaching the eye. As it emerges from the drop of water the white light has been split into an array of colours.

Only three manuscripts of Theodoric’s short tract on the rainbow survive, and we know of only one medieval discussion of his discovery.³⁷ Regiomontanus, it is true, planned to publish him, but, where other texts Regiomontanus had planned to publish appeared in print in due course, Theodoric’s little treatise did not.³⁸ In 1514 a summary of his argument was presented in a physics textbook intended for students at Erfurt (and an even shorter one appeared in 1517, this time without any illustrations).³⁹ There is no evidence that these summaries had any influence. Theoderic’s work then disappears entirely from view until it was rediscovered in the nineteenth century. When Descartes produced his study of the rainbow he had to start from scratch, despite the fact that he was very largely simply repeating the work both of Theodoric and of al-Farisi.⁴⁰ Thus it is important to see that when we hail Theodoric as an important scientist our judgement is essentially anachronistic: he did not seem important to his contemporaries or successors, and his influence is negligible. His work would have been much more likely to have been preserved and copied if it had taken the form of a commentary on Aristotle’s Meteorology – the most widely read commentary was that of Themo Judaei, which contains no reference to Theodoric – and if it had not depended on elaborate illustrations which were difficult to copy accurately.

A similar point may be made about the work of Ibn al-Haytham. Only one complete manuscript of the original Arabic text of The Optics survives (the vast majority of Ibn al-Haytham’s work – he wrote two hundred texts – has been lost), and the only Arabic commentary on his optical work that we know of (before the modern day) is that of al-Farisi (1309).⁴¹ Ibn al-Haytham was much more widely discussed in the Latin West than the Muslim East, but even in the West he was treated as a text, not as a handbook of experimental practice. No one, as far as we know, replicated his experiments. Thus in both Arabic and medieval culture experimentation had an uncertain status: it existed, but it was not admired or imitated. It was recognized as a form of knowledge, but only at the margins. In both cultures Ibn al-Haytham was seen as a model to be imitated only when it came to looking for an explanation for the rainbow; for the vast majority of medieval authors learning was something to be found in books and to be tested by abstract reasoning; it was not to be found in things and tested by experimentation.

§ 4

Thus there was nothing unprecedented about experimentation in 1648; there were good precedents, one of which (Ibn al-Haytham’s Optics) was widely known, if rarely imitated. Rather, the significance and status of experimental knowledge underwent a peculiar transformation in the course of the seventeenth century. It moved from the margins to the centre.xixi Kant claimed that the experimental method of the seventeenth century (he cites Galileo, Torricelli and the chemist Georg Stahl) represented ‘the sudden outcome of an intellectual revolution’, the moment when natural science entered on ‘the highway of science’; this judgement is sound not if it is understood as claiming that Galileo and Torricelli were the first to conduct experiments but, rather, read as maintaining that previously experiments were seen as no more than a byway.⁴² Above all, experimentation began to engage directly with central claims made by Aristotle. At the same time, those who performed experiments ceased to be lonely and isolated individuals; they became members of an experimental network. Why exactly did the significance and status of experimental knowledge change? We need to look more closely to understand what happened.

The first major field for experimental enquiry in the early modern period was the magnet, a subject on which there was virtually no classical commentary (because the compass was unknown in antiquity), which meant that the experimental approach faced fewer obstacles than with any other topic. Moreover, the importance of the compass in navigation meant that the magnet was bound to be a subject of discussion. The first attempt at an experimental study of the magnet was reported in Pierre de Maricourt’s Letter on the Magnet (1269); Pierre describes the polarity of magnets, shows that like poles repel each other and unlike poles attract each other, and describes how iron can be magnetized. Despite surviving in thirty-nine manuscript copies, there is no evidence that it led to further experimental work until it finally appeared in print in 1558.⁴³ Like Ibn al-Haytham, like Theodoric, Pierre de Maricourt had no immediate successors.

By 1522 Sebastian Cabot had discovered the variation of the compass: the needle does not point to true north but either somewhat to the east or to the west, and it varies in the extent to which it diverges from true north depending on where you are on the surface of the globe. This discovery presented fundamental difficulties for any account of how the compass worked, but it also raised the exciting possibility that variation might be sufficiently regular for it to be used to measure longitude. Since longitude was the missing piece of knowledge for oceanic navigators, all the early modern studies of the magnet were alert to the possibility that they might be able to fill this gap.

Chronologically speaking, Leonardo Garzoni’s treatises (discussed in Chapter 7) are the first major work of modern experimental science, but here chronology is misleading, for in crucial respects they are simply a continuation of the erratic medieval tradition of experimentation. Their conceptual apparatus is Aristotelian, and they seek to address a gap or anomaly in the Aristotelian scheme of knowledge. They respond to the age of exploration and discovery, but only with the intention of preserving and protecting the conceptual apparatus of traditional philosophy. And, just like the medieval experimentalists before him, Garzoni had almost no impact. As far as his colleagues were concerned, his work was of marginal interest unless it could be shown to result in a method for identifying longitude. Only one copy of his manuscript survives, and later Jesuit theorists rediscovered his work only because they needed ammunition to use against William Gilbert. It is true, as we have seen, that Garzoni’s treatises were picked up by della Porta, who stole material wholesale from them, and who at least tested his claims about garlic and diamonds, but this was because the subject of magnetism, which involved hidden and inexplicable forces, fell squarely within the territory of natural magic; it is not because della Porta was converted by Garzoni to a new way of thinking, or a new, more reliable experimental practice.

Gilbert’s On the Magnet takes us into a different world; indeed, he claims to be engaged in a new type of philosophizing. (It is an interesting question whether he would have made this assertion with the same confidence if he had read Garzoni.) For Gilbert, the experimental method is an alternative, not a supplement, to Aristotle. The goal of his philosophy is to make new discoveries, not to patch and mend an existing body of knowledge. De Maricourt and della Porta were crucial sources for Gilbert: we can tell that he repeated their experiments with care, which is how he knew that della Porta was copying from a half-understood source. He also had the advantage that the dip of the compass (that is, its tendency to point downwards from the horizontal by different amounts in different places) had been discovered by Robert Norman in 1581.⁴⁴ Gilbert was the first to recognize that the Earth itself is a magnet, and that this is why the compass needle points to the north. Others before him, including Digges, had grasped that the compass needle was not attracted towards a particular location, either in the heavens or within the Earth, but they had not taken the further step of thinking of the whole Earth as a magnet.⁴⁵ But this was not the limit of Gilbert’s ambition. He sought to show (building on a suggestion by de Maricourt) that a magnet has a natural tendency to rotate on its axis; this, he claimed, provided an explanation for at least one of the three movements attributed to the Earth by Copernicus. Thus Gilbert made magnetic experiments relevant to a well-established branch of natural knowledge, astronomy. But at the same time he failed in his ultimate objective: he could not produce an experiment in which his magnets spontaneously rotated.

Gilbert’s Copernicanism guaranteed a hostile response from orthodox Catholic scholars once Copernicanism had been condemned in 1616.⁴⁶ Thus Niccolò Cabeo, who was largely reproducing the arguments of Garzoni, continued to insist that there were two separate phenomena, the attraction of iron by magnets and the tendency of magnets to point towards the pole, not, as Gilbert claimed, one underlying phenomenon to which both could be reduced. It also guaranteed a favourable response from Copernicans such as Galileo and Kepler: Galileo said his own method was rather like Gilbert’s, and Kepler took Gilbert’s account of magnetism as a model for the sort of forces which might drive the planets in orbit around the sun. But, after Gilbert, work on the magnet failed to discover the regularities that make experimental knowledge possible. Not only did variation and dip diverge from place to place, but in 1634 a group of English experimenters claimed that variation fluctuated over time. This depended on their confidence in the reliability of measurements taken decades apart; others were quick to dismiss such findings as the result of defective technique. Nature could not be so capricious. Eventually, however, the naysayers were forced to concede that not only did variation change over time, but so did dip.⁴⁷ If successful experimentation depends on nature being economical and regular, then the study of the magnet after Gilbert seemed to undermine the conviction that this was so.

What do de Maricourt, Norman, Garzoni, della Porta and Gilbert have in common? Essentially, nothing. De Maricourt was a mathematical scholar and, it would seem, a soldier. Norman was a sailor who took advice from men of learning. Garzoni was a Jesuit, a Venetian patrician and a scholastic philosopher. Della Porta was a Neapolitan nobleman who had made a profession out of occult learning. Gilbert was an English doctor and the advocate of a new philosophy. Della Porta was preoccupied with sympathy and antipathy, while Garzoni and Gilbert refused to use such categories. This ‘nothing’ is important because it undermines the standard account of the origins of experimental science. It will not do to claim, as Marxists do, that sixteenth- and seventeenth-century experimentation involved a new collaboration between intellectuals and artisans when already, in 1269, de Maricourt had said that anyone studying the magnet must be ‘very diligent in the use of his own hands’, so manual dexterity was not new in the sixteenth century; and there is nothing to suggest that Garzoni, who certainly was dextrous, had links to the world of the skilled craftsman.⁴⁸ The study of variation and dip evidently depended on collaboration between navigators and intellectuals, but so did the whole science of cartography. It leaves us at the same time with a problem. If Gilbert is an example of a new type of scientist, what makes this new science possible?

The compass allows you to navigate out of sight of land and, naturally, Edward Wright’s prefatory letter to On the Magnet mentions the circumnavigations of the Earth by English sailors. But in his preface Gilbert presents himself as sailing on a quite different ocean, an ocean of books. And, indeed, he had either bought books in vast quantities or had had access to a remarkable library, for On the Magnet begins with the first systematic literature review. Gilbert had read everything that had ever been written on the magnet. No ancient or medieval author (at least not since the great library of Alexandria went up in flames in 48 BCE) could have done this. Gilbert can confidently declare that he has made new discoveries because he knows exactly what was known before. He insists that knowledge comes not from books only, but from the study of things; yet the simple truth is that the ocean of books is as important to his researches as are the oceans.

So we might want to claim that it is the book – or rather, in this case, the well-stocked library – that transforms the status of experiment; by crystallizing past knowledge, the library makes new knowledge possible. In the case of anatomy, a single book, Vesalius’s Fabric of the Human Body, had functioned as a whole library, but each new field required a similar enterprise of assimilating existing knowledge before new discoveries could begin. Printing made facts, as we saw in Chapter 7; and it begins to look for a moment as if it may also have made the new experimental philosophy.

But what is important to Gilbert is not just book learning, or even the performing of experiments. There is a third element to his new science. He thanks

[s]ome learned men … who in the course of long voyages have observed the differences of magnetick variation: the most scholarly Thomas Hariot, Robert Hues, Edward Wright, Abraham Kendall, all Englishmen; Others there are who have invented and produced magnetical instruments, and ready methods of observation, indispensable for sailors and to those travelling afar: as William Borough in his little book on The Variation of the Compass or Magneticall Needle, William Barlowe in his Supply, Robert Norman in his Newe Attractive. And this is that Robert Norman (a skilful seaman and ingenious artificer) who first discovered the declination [i.e. dip] of the magnetick needle.⁴⁹

Gilbert thus acknowledges a little community of experts, many of whom are known to him personally (Harriot, Borough and Norman for example; Edward Wright was a close collaborator). Where all previous experimenters, from Galen to Garzoni, appear to have worked in isolation, we have here for the first time a functioning scientific community, and the quality of Gilbert’s work depends in part on his belonging to this community. There is no doubt that the later discovery of the variation of variation depended upon having a close-knit community of experts using the same instruments and techniques who acknowledged the accuracy of each other’s measurements over long periods of time.

§ 5

It would be difficult to overestimate the impact of Gilbert’s On the Magnet, not because everyone was interested in magnets but because for the first time the experimental method had been presented as one capable of taking over from traditional philosophical enquiry and transforming philosophy. Central to Gilbert’s enterprise was the claim that you could reproduce his experiments and confirm his results: his book was, in effect, a collection of experimental recipes. In Padua in 1608 Galileo copied Gilbert’s technique for arming a magnet by coiling iron wire around it, used it (without acknowledging his debt to Gilbert) to create what he claimed was the strongest magnet in the world and promptly sold his super-magnet for a lot of money to the Grand Duke of Florence.⁵⁰ Others, too, were surely copying and testing Gilbert’s experiments, although no one else seems to have found a way of making money out of doing so. It is worth noting, then, that there is no record of anyone ever having claimed that Gilbert’s experimental results could not be replicated. Replication is a vexed issue in the history of science but, where the history of magnetism is concerned, matters are straightforward: good results can be replicated, and bad ones (which would include some of Garzoni’s) cannot.

There can be no replication if there is not some form of publication, or, at least, communication. Two scientists discovered the law governing the acceleration of falling bodies at around the same time – Harriot and Galileo.⁵¹ Harriot kept his results to himself; Galileo finally published in 1632, some decades after he had made his discovery. He claimed that, leaving air resistance aside, heavy objects and light objects would fall at the same speed, so that if one dropped a musket ball and a cannon ball, or a wooden ball and a lead ball, of the same size, simultaneously from a tall building, they would hit the ground at the same moment. Soon all sorts of people were dropping objects off tall buildings and getting rather varying results (it is actually much harder than one might think to drop two object simultaneously, and very difficult to measure how far apart they are when the first one hits the ground). In France, Marin Mersenne went to elaborate lengths in 1633 to replicate Galileo’s experiments and conduct accurate measurements. Where orthodox Aristotelianism said that objects fell at a constant speed, and the heavier they were the faster they fell, Galileo maintained that they accelerated as they fell, and that they all did so according to the same principle. His claims provoked a widespread concern with the replication of his experiments precisely because the viability of Aristotelian orthodoxy was now at stake.⁵²

In 1638 Galileo also published the claim that if a column of water in a suction pump or in a tube with an air-tight seal exceeded a certain height (32 feet, he said), the column would descend, leaving a vacuum above it. He thought (rightly) that the key issue was the weight of the water: just as a dangling rope would break under its own weight if it was long enough, so at a certain point a column of water would break. What held the column of water together was a natural force, the resistance to the creation of a vacuum, and this force was a major factor in understanding the strength of materials. This was contrary to orthodox Aristotelian philosophy, which said that there could be no such thing as a vacuum.

Galileo had clung to his account despite the fact that a friend, Giovanni Battista Baliani, had suggested an alternative explanation for the fact that suction pumps ceased to work if they were asked to lift water more than 18 braccia (roughly 35 feet) – a figure established by experience. The explanation, Baliani said, was that up to a certain point the weight of the water was balanced by the weight of the air pressing down on all of us all the time; above that point the column of water could not rise and, in a pump without leaks, a vacuum would be created. There was no ‘resistance’ to the creation of a vacuum.⁵³ Either way, Galileo’s and Baliani’s claims struck at the very heart of Aristotelian physics. What Gilbert had aspired to do, Galileo had actually done: he had announced discoveries which were entirely at odds with the established philosophy. At first, a guerrilla army picks off enemy outposts and interferes with communications, but then, if it gathers strength, it must eventually move from hit-and-run attacks to a massed engagement with the enemy. Galileo’s Dialogue Concerning the Two Chief Systems was a full-scale engagement with traditional astronomy; his Two New Sciences (1638) represented a full-scale attack on Aristotelian physics. Dispute over Copernicanism was distorted by the intervention of the theologians, but the dispute over physics could take place without such interference. Battle was engaged.

In Rome, sometime after 1638, a group of orthodox philosophers set out to prove that Galileo was wrong about the vacuum. Gasparo Berti built a long lead tube with a window in one end, filled it with water and sealed it at both ends, then upended it in a barrel of water and removed the seal from the bottom. At first, no empty space appeared at the top, but then he realized that he needed to measure the height of the column not from the ground but from the top of the water in the barrel and raised the tube a little higher; at once the column of water descended and an empty space appeared. But was it actually empty? Light travelled through it. A bell was fitted into the top of the tube, and it could be made to ring, so it seemed there was air present. (The vibrations of the bell must have been carried by the strut supporting it, rather than by the air.) The results were inconclusive. They had neither refuted nor confirmed Galileo’s claim; they had simply produced an anomaly. And so the philosophers turned their minds to other things; later, no one could even remember the year in which they had performed this experiment, and no one wrote about it at the time.

In Florence, however, Galileo’s disciple Torricelli heard in 1643 of Berti’s experiments and realized that he could simplify matters by using a denser liquid. A tube containing mercury would need to be one-fourteenth the height of a tube containing water: if 32 feet of water was the crucial height in order to generate a vacuum, then only a little over 2 feet of mercury would be needed. So he repeated the experiment with mercury, and reproduced the anomalous space. He had reached the same conclusion as Baliani: the space contained a vacuum, and the weight of the mercury was balanced by the weight of the air. We live, he wrote, under an ocean of air. Since the air on some days seemed to be heavier than on others, he reasoned that he ought to be able to measure the changing weight of the air. But his barometer produced puzzling and inconsistent results (it had probably been damp when he introduced the mercury), and he put it aside – and then died before he could learn of the successes of others.⁵⁴

This is Schott’s representation (some twenty-five years after the event) of the first attempt to create a vacuum at the top of a tall tube filled with liquid, in an attempt to disprove Galileo’s claim that a vacuum would appear if the column of water was more than about 11 metres high. The space at the top of the tube has been expanded to include a bell, on the theory that if there was a vacuum no sound would be audible. Gaspare Berti’s experiments were inconclusive (a sound was heard from the bell, suggesting there was no vacuum), but provided the inspiration for Torricelli’s decision to substitute mercury for water. Berti’s experiment was first described in print in Niccolò Zucchi’s Experimenta vulgata (1648).

In France, Mersenne received a rather garbled account of Torricelli’s experiment and tried unsuccessfully to reproduce it, but he lacked the right sort of glass tube. Shortly afterwards, toward the end of 1644, he travelled to Florence, where he met Torricelli, and then to Rome, where he may have seen Torricelli’s experiment; on his return to France he tried unsuccessfully to reproduce the experiment, but his glass tubing was not of high enough quality. In the autumn of 1646 Pierre Petit and his friend Blaise Pascal successfully performed the experiment in Rouen; Petit had heard about it, but neither had seen it done before. Pascal then reinvented (for he had heard nothing of it) the experiment originally performed in Rome, substituting red wine for water to make it easier to see the result. Berti’s experiment had been performed in a public place, but there is no evidence that it attracted any attention. Pascal’s experiment was different: it was put on as a public show, but still there is no reason to think that Pascal intended to publish any time soon. When others, however, began to debate the significance of what he had done he rushed his own account into print (in 1647) in order to establish his claim to priority. Pascal sent copies of his booklet to all his friends in Paris and to every town in France where he thought there were people who might be interested in reading it – presumably, to the local booksellers, for somewhere between fifteen and thirty copies went to Clermont-Ferrand alone; Mersenne sent copies to Sweden, Poland, Germany, Italy, and indeed all over the place. The status of experimentation was changing; and Pascal and Mersenne made every effort to bring this about.⁵⁵

According to Vincenzo Viviani, the young Galileo in around 1590 had dropped objects off the Tower of Pisa and the whole university had gathered to watch. Viviani was probably right to say that Galileo performed this experiment, and if he did he was not the first: similar experiments had been performed by Giuseppe Moletti in 1576 (but never published) and by Simon Stevin (who published in 1586, but attracted no attention, partly because he wrote in Dutch).⁵⁶ But there is absolutely no evidence, other than Viviani’s account much later, to suggest that crowds gathered to watch Galileo’s early experiments. In assuming they did, Viviani is reading back into Galileo’s youth a view about the status of experiments which became established only in the 1630s: Viviani became Galileo’s assistant in 1639 at the age of seventeen, and wrote Galileo’s Life in 1654. Pascal’s experiments of 1646, on the other hand, really did draw crowds.

Up to this point, the story of the vacuum experiments is a story of accidents and near-misses. Berti’s experiment had reached no clear conclusions; Baliani and Torricelli had got the theory right, but Torricelli’s experiment had not quite worked, and the initial reports that had reached France had not conveyed his theory, nor provided sufficient detail to enable the experiment to be reproduced. From 1646 Torricelli’s experiment was widely performed, even though mercury was expensive and it continued to be difficult to get sufficiently sturdy long tubes sealed at one end. In fact, Torricelli’s experiment rapidly became famous: the phrase ‘famous experiment’ is first used in English in 1654 to refer to it, and for an Italian author in 1663 it is famosissima.⁵⁷

Once Torricelli’s experiment had been accepted as a basic model, it was possible to devise all sorts of variations. Three are of great importance. First, Pascal invented a way of putting a barometer within the anomalous space at the top of the Torricellian tube: when the mercury dropped in the main tube to a height of 27 inches, it dropped to zero (or very near zero) in the second barometer within the Torricellian space. Various revisions and improvements on this experiment were devised by Pascal and others; in its different forms, the ‘void in the void’ experiment seemed to confirm that there was no (or almost no) air pressure within the Torricellian tube. Second, Pascal devised the Puy-de-Dôme experiment. Third, Roberval devised an experiment where a carp’s bladder which had been flattened and sealed tight was placed at the top of the Torricellian tube. When the mercury dropped, the bladder was left behind and swelled up as if it had been pumped full of air. The interpretation of this experiment was far from straightforward, but Roberval argued that if there had been air in the carp’s bladder when it seemed there was none, so, too, could there be air in the Torricellian space (although only in the most minute quantities).⁵⁸

In thinking about the expansion or rarefaction of air, Roberval invented the concept of ‘the spring of the air’, which Boyle went on to make famous in his New Experiments of 1660, and which became systematized in Boyle’s law (1662).⁵⁹ It is true that Roberval did not use the word ‘spring’, borrowing the Latin word elater (which shares a root with ‘elastic’; elater is the translation for ‘spring’ in the Latin version of the New Experiments) from Mersenne;⁶⁰ but Boyle’s failure to acknowledge any debt to Roberval for the concept shows that intellectual property in theories was slower to become established than intellectual property in other sorts of discovery, such as the design of an experiment. Already in 1662, however, Boyle was careful to acknowledge that a number of people had contributed to the formulation of Boyle’s law.⁶¹ And the concept of intellectual property certainly had been established by 1677, when Oldenburg, writing in the Philosophical Transactions, complained that a Latin translation of Boyle’s works, produced in Geneva without his permission, failed to record the date when the originals were first published, which could give the false impression that Boyle had stolen from others when they in fact had stolen from him. In the Second Continuation Boyle, or rather his publisher on his behalf, returned to the issue: ‘[F]or though some Writers have with Ingenuity enough cited the Name of our Author in their Works, yet more have done otherwise, transferring not a few of his Experiments, together with the Ratiocinations explaining them, after the manner of Plagiaries into their Books, making no mention of his Name at all.’⁶² A few years earlier, the Chief Justice, Matthew Hale, writing anonymously on the Torricellian experiments, had been anxious to insist that he had cited his sources, in order to ‘avoid, as much as I can, the imputation of a Plagiary’.⁶³

That someone who originates a new idea has a right to be acknowledged seems to us obvious, but this idea was fundamentally new. If we look back to the Parisian philosophers of the fourteenth century, for example, to Oresme, Buridan, John of Saxony and Pierre d’Ailly, we find ourselves in a world where scholars reported each others’ arguments but failed to record who originated any particular line of argument, so that historians still cannot write the history of the school of Paris in terms of who influenced whom; being first was not what mattered to fourteenth-century philosophers. This world still existed in 1629 when Niccolò Cabeo published his Magnetical Philosophy, which is almost entirely drawn, and indeed much of it taken verbatim, without acknowledgement, from Leonardo Garzoni’s unpublished manuscript; it still existed in 1654, when Pascal completed his Treatise on the Equilibrium of Liquids: in it he drew heavily on works by Stevin, Benedetti, Galileo, Torricelli, Descartes and Mersenne, but made no mention of any of his predecessors.⁶⁴ It still existed in 1660, when Boyle (who had a highly developed sense of propriety and surely had no consciousness of doing wrong) borrowed without acknowledgement from Roberval, but was fast disappearing by 1682, when he complained about the borrowings (still, perhaps, entirely innocent) of others. In 1687 Boyle’s friend David Abercromby announced his intention of writing a treatise which would be a history of discovery through the ages – the book that Polydore Vergil had failed to write. Included would be all ‘new Contrivances, whether Notions, Engines, or Experiments’. This would be a study of what he calls ‘authors’, that is, discoverers and inventors: ‘By Authors, here are meant, those that are really such [as opposed to plagiaries or those responsible for merely incremental improvements], and the first Inventors of any useful piece of Knowledge’.⁶⁵

a) Adrien Auzout’s void-in-the-void experiment, from Jean Pecquet, Experimenta nova anatomica (1651). In this experiment, a barometer inserted into the Torricellian void at the top of a first barometer measures the air pressure there: the mercury in the second tube does not rise, marking the absence of any air pressure in the space; when air is introduced into the space at the top of the first barometer the mercury in that barometer drops down into the tank, while the mercury in the second barometer rises to a height of 27 inches.

b) Gilles de Roberval’s carp-bladder experiment. A carp bladder, from which all the air has been squeezed out, and which has been tied off, is introduced into the Torricellian void. It promptly swells up, demonstrating the extraordinary elasticity of the little bit of air left in the bladder. Roberval thought this justified the conclusion that there is always some air, however little, in the Torricellian void.

From 1646 to 1648 a small group of experimenters (Pascal, Roberval, Auzoult, Petit, Périer, Gassendi, Pecquet) scattered across France were working on vacuum experiments simultaneously. What held them together was their common friendship with Mersenne, with whom they exchanged letters and at whose home they met when they were in Paris. They had a variety of professional commitments, but they regarded themselves as first and foremost mathematicians, and many of them made major contributions to pure mathematics.⁶⁶ They competed with each other and collaborated, and (for the most part) trusted each other enough to be confident that their own contributions would be recognized. They freely circulated manuscripts among themselves. Roberval, for example, never published his vacuum experiments, but a letter he wrote describing the early history of the experimental programme in France was published in Poland, a number of his experiments were described in print by an opponent, and his carp’s-bladder experiment was published by Pecquet in 1651 in a volume primarily devoted to new anatomical studies (this was translated from Latin into English in 1653). Publication was important within this group, but no more important than private and semi-public correspondence: Mersenne wrote letters to Italy, Poland, Sweden and Holland announcing Pascal’s Puy-de-Dôme experiment.⁶⁷ It is also significant that Mersenne’s friends collaborated without agreeing with each other. There was agreement on the value of experimental enquiry, not on how to interpret the results.

Mersenne died in 1648, and little further progress in research on the vacuum was made in France. But Pascal’s and Pecquet’s works were read in England (where Henry Power immediately replicated Pecquet’s experiments) and Italy, as was Gaspar Schott’s Mechanica (1657). Schott reported not only Berti’s original experiments in Rome but also von Guericke’s construction of a vacuum pump.⁶⁸ Von Guericke had demonstrated how two hemispheres from which the air had been extracted were held so tightly together by air pressure that teams of horses could not pull them apart. It was Schott’s book that inspired Robert Boyle to construct his own air pump in England, and it is perhaps not a coincidence that vacuum experiments were recommenced in Florence in 1657. If one drew up a list of all the people known to have performed barometer experiments between Torricelli in 1643 and the discovery of Boyle’s law in 1662, it would reach a hundred names without too much difficulty. These hundred people are the first dispersed community of experimental scientists.xiixii

Experiments produce new knowledge, but if that knowledge does not circulate there is little opportunity for further progress. Torricelli’s barometer represents the first piece of experimental apparatus that became standardized and widely available; and endless variations on the experiments that could be performed with it (such as releasing insects into the Torricellian space) were devised. This was the first time experimentation had had an audience (symbolized by the small crowd surrounding Périer on the summit of the Puy-de-Dôme) and it was the first time it became a collaborative and competitive process.

As one would expect, this first successful experimental community changed the way in which scientific communities were constructed and what they were used for. Mersenne’s community was an informal group which met and exchanged letters, though he expressed a wish to form a proper college that would function mainly through correspondence. There had been earlier semi-scientific academies: della Porta had formed an academy (which the Inquisition had forced him to close) dedicated to the pursuit of secret knowledge, while both he and Galileo had belonged to the Accademia dei Lincei (the lynx-eyed) founded by Prince Cesi.xiiixiii Bacon had imagined a scientific community at work in his utopian New Atlantis (1626). Mersenne was not the first, and nor would he be the last, to establish an ‘invisible college’ through correspondence: his own network grew out of that built up by Peiresc, and similar networks were founded by Hartlib and Oldenburg in England (Oldenburg’s network becoming identified with the Royal Society when he became, jointly with Wilkins, its first secretary).⁶⁹

Schott’s representation of the Magdeburg hemispheres in Experimenta nova (1672). In 1654 in Regensburg, and then again in 1656 in Magdeburg, Otto von Guericke evacuated the air from a copper sphere with an air pump. He then attached teams of horses to the sphere, which consisted simply of two hemispheres mated together, but they were unable to draw them apart. This demonstrated the force of air pressure acting on the hemispheres and inspired Boyle’s construction of an air pump.

The extraordinary success of the Torricellian network, as we may term the group of people involved in barometer experiments, was a major factor in leading to the establishment of the Accademia del Cimento in Florence (1657), the Académie de Montmor in France (1657), the Royal Society in England (1660) and the Académie Royale in France (1666). The Accademia del Cimento published a single book, but the Royal Society published the first journal, the Philosophical Transactions (from 1665), devoted to the new science. In France the Journal des sçavans began publishing the same year: it covered a wide range of academic subjects but declared in its first issue that one of its major concerns would be to announce new discoveries.

Thus the informal Torricellian network marks the effective beginning of the institutionalization of science, driven by the conviction that collaboration and exchange would lead to more rapid progress. As one would expect, this was accompanied by a new commitment to the idea of scientific progress. In a draft preface to an unpublished book on the vacuum (c.1651), Pascal distinguished between forms of knowledge that were historical in character and depended on the authority of the sources on which they relied (theology was the key example), and forms of knowledge that depended on experience. In the case of the latter, each generation, he claimed, knew more than the one before, so progress was continuous and uninterrupted (‘all mankind continually makes progress as the world grows older’).⁷⁰ Pascal says, indeed, that each generation sees further than the one before. He almost certainly has in mind the famous saying that we are dwarfs standing on the shoulders of giants. The saying originates with Bernard of Chartres in the twelfth century, but is now usually quoted from one of Newton’s letters: ‘If I have seen further it is by standing on the shoulders of giants.’ Newton was engaging in false modesty (and the shoulders he had been standing on most immediately were those of Hooke, who was distinctly short of stature).⁷¹ Since Pascal’s purpose was to undermine respect for antiquity, he had no intention of repeating the claim that the ancients were giants as compared to us; he assumes that each generation is equal to any other in its abilities.

Boyle’s first air pump, designed and made by Robert Hooke, from Boyle’s New Experiments Physico-mechanical (1660).

The frontispiece of the English translation of the experiments of the Accademia del Cimento: Nature is shown turning her back on Aristotle and being introduced by the Accademia to the Royal Society. Published in Italian in 1666, the Saggi were presented to the Royal Society (they were published in a luxury edition intended not for sale but for presentation); after considerable delay Richard Waller translated them as Essayes of Natural Experiments (1684). The frontispiece was drawn by Waller himself.

At first sight, Pascal’s claim that progress has been continuous seems nonsense: we would accumulate experience from generation to generation only if we had a reliable way of recording and transmitting it. Pascal is, however, assuming a book and not a manuscript culture (we are back with Gilbert’s ‘ocean of books’); it is only since the invention of printing that knowledge has been effectively recorded and transmitted. Moreover, he recognizes that not every individual makes progress; rather, it is what he calls l’homme universel, human beings collectively. It was through his collaboration with other scientists that Pascal came to have the sense of belonging to a collectivity greater than himself. The Torricellian network solved problems more efficiently than any individual could on their own. As history, the preface is bunk, because in 1651 progress was new. But since then it has indeed been uninterrupted and continuous; as an account of modern science, the preface is spot on.

Thus experiments are not new. The first person who rubbed two sticks together to make a fire was conducting an experiment. Galen, Ibn al-Haytham and Theodoric of Freiberg performed experiments. What is new is the scientific community that is interested in experiments. We can see a foreshadowing of this in the group of people which surrounded Gilbert as he conducted experiments on magnets, but for the most part these were fairly uneducated navigators. It takes proper form in the years after the publication of Galileo’s Dialogue Concerning the Two Chief World Systems in 1632. Even Daryn Lehoux, who thinks the Romans had everything we have, acknowledges one exception:

There were no ancient universities, no scientific conferences, no journals where investigators published their results. So, too, no New Scientist, no science pages in the New York Times where the newest work could be reported, compared, commented on. From these modern sources there often emerges an understanding, among professionals and among the scientifically literate public, of something we might call the ‘consensus in the field’ on many issues.⁷²

Kuhn had a particular term for ‘the consensus in the field’. He called it ‘normal science’, as opposed to revolutionary science. The Torricellian barometer was the first experimental apparatus around which a normal science developed. There had been stable, consensus-based sciences before: Ptolemaic astronomy, for example, or Vesalian anatomy. But this is the first time a consensus developed around what the English call an ‘experiment’.

In the course of the seventeenth century the Latin words experientia and experimentum and, with them, the English words ‘experience’ and ‘experiment’ began to diverge in meaning. Thus from 1660 on ‘experimental philosophy’ became a widely used label for a science relying on experiments; no one wrote of an ‘experiential philosophy’.⁷³ The divergence has its roots back in the early thirteenth century, when translators of key Arabic texts, such as Ibn al-Haytham’s Optics, had chosen the Latin experimentare rather than experiri to translate the Arabic i’tibar and to describe experiments in optics.⁷⁴ Experimentum was, as a consequence, the word generally used by medieval philosophers to describe artificially constructed experiences. It will have been the obvious choice for Gilbert in his On the Magnet. Slowly, ‘experiment’ became in English a technical term for something scientists do; but not, as we have seen, in French. In Italy, Galileo normally wrote of esperienza in Italian, when in Latin he would have written of experimentum. Esperimento and sperimentare were neologisms and, though they were to be found in the dictionary of the Accademia della Crusca (1612), they were not used in any of the classic texts of Tuscan literature. But esperienza was too broad a term to identify the procedures of the new science, and after Galileo’s death his disciples formed the Accademia del Cimento (cimento means ‘test’ or ‘proof’ in the practical sense, like the English word ‘assay’ and the French word essai, so this was an academy devoted to experiments). The ultimate success of esperimento reflects, just like the phrase philosophie expérimentale in French, the influence of the English language and of English science. In English, the phrase the ‘experimental method’ first appears in 1675.xivxiv

Thus, in the case of the word ‘experiment’, linguistic change lags behind both theory and practice. If language provides only very limited assistance, how are we to know an experiment when we see one? The answer is simple: an experiment is an artificial test designed to answer a question. The Latin term for this, well known to medieval and Renaissance philosophers, is periculum facere, to make a test or trial of something.⁷⁵ Such a test usually involves controlled conditions, and often requires special equipment.

§ 6

Up to this point, when I have tried to identify what it is that we have that they (Greek, Roman and medieval philosophers) didn’t, my response has been to locate a conceptual tool, such as discovery or the fact, or a technical breakthrough, such as measuring the non-parallax of comets, or an instrument, such as the telescope. This chapter has pointed instead to a sociological reality – the scientific network – and more concretely, the small crowd that surrounded Périer on the summit of the Puy-de-Dôme. It would be wrong to overemphasize the difference between conceptual and sociological explanations: discoveries have to be announced, facts accepted, experiments replicated; the concepts are grounded in a sociological reality – the audience (created, above all, by the printing press). The scientific network is another term for that sociological reality: Pascal announced his discoveries to Mersenne’s network, and persuaded them that his facts were right by getting them to repeat his experiments. The new concepts and the new social organization are two sides of one coin. If earlier scientists, like Harriot in England, failed to publish, or, like Galileo in Italy, at least as far as his new physics was concerned, were slow to publish, it is partly because they were not confident there was an audience for what they had to say. The success of the Torricellian barometer created an audience for the new science.

In stressing that science is a community activity I do not mean to imply (any more than Kuhn did) that science only has a social history, or (as relativists have it) that science is whatever scientists agree on. There had been earlier attempts to build communities whose goal was to advance knowledge: sixteenth-century doctors, for example, had formed networks and had published the resulting correspondence.⁷⁶ What such communities had never been able to do, however, was to build consensus around the problems to be worked on and the solutions that would be regarded as satisfactory. They never established anything resembling normal science. The key to normal science is replication. Over and over, in the years after 1647, scientists filled long glass tubes, sealed at one end, with mercury and inverted them in baths of mercury. They provided each other with helpful guidance: don’t even breathe in the tube, says Pierre Petit, or you will contaminate the mercury with water; put the apparatus on a blanket, says Henry Power, and have a wooden spoon to hand, so that any mercury you spill is captured and you can scoop it up.⁷⁷ They invented endless variations: but nobody who performed a variation had not also performed the basic experiment. And over and over they got the same results.xvxv If the Torricellian barometer had not been easy to replicate, it would never have become the first famous experiment. When the Accademia del Cimento was formed in 1657 its motto was a tag from Dante, provando e riprovando, and they tested and tested again. Successful replication (not intellectual coherence or the support of authority) was now the mark of reliable knowledge.

I am here arguing against a powerful tradition in recent historiography of science which insists that replication is always problematic, and that in the end what counts as successful replication is always decided by the intervention of authority.⁷⁸ According to these historians, replication is a social artefact, not a natural fact. The classic study is Leviathan and the Air-pump, by Steven Shapin and Simon Schaffer.⁷⁹ That book, which has been described as the most influential work in the history of science after Kuhn’s Structure of Scientific Revolutions, presents a number of arguments which have become famous.⁸⁰ It argues that Boyle, through his air-pump experiments, was a pioneer in making facts: from the previous chapter we can see that this view is mistaken, unless one focusses narrowly on the use of the word ‘fact’. Gilbert, Kepler and Pascal all established facts. The book also contends that Boyle developed a novel technique for winning support by turning readers into virtual witnesses: virtual witnessing is important, but Boyle was hardly a pioneer in this respect either. It maintains that the dispute between Boyle, who claimed (albeit cautiously) to have made a vacuum, and his opponents, who claimed that he had not, was not resolved because Boyle had the better arguments but because he had the more powerful social position.

Here it is important to compare the disputes in which Boyle was engaged with those in which Pascal was engaged. Boyle constructed his air pump because the glass globe out of which the air was evacuated provided a better site for experiments than the space at the top of a barometer. Boyle, for example, could place a lit candle in his globe, or a bird; although insects and frogs could be passed through the mercury into the Torricellian space, flames and birds could not. In order to demonstrate that his experimental space was equivalent to that in a barometer, Boyle had to repeat the standard experiments, such as the void in the void and the carp’s bladder, and show that he obtained the same (or virtually the same) results. In so far as Boyle’s ‘vacuum’ was indistinguishable from Torricelli’s, the arguments in England were fundamentally the same as those which had already taken place in France. Against those who maintained that, since light was able to pass through the Torricellian space, there must be within it some mysterious aether, a substance which appeared to have no weight and was present everywhere, Pascal had replied that, as the nature of light was unknown, it was futile to insist that it required some imaginary substance as its medium for transmission. To claim that there existed a substance with no measurable attributes was, he maintained, to turn physics into a fantastical story, like Don Quixote. Pascal and Boyle both ‘won’ this argument in so far as they succeeded in casting doubt on the legitimacy of appealing to substances whose existence might be theoretically convenient but could not be demonstrated experimentally.

However, a standard argument against Boyle’s experiments with his air pump was that the pump leaked, and that therefore it was incapable of making a vacuum. This was true, and in the void-in-the-void experiment the mercury never fell below half an inch. The Torricellian barometer, by contrast, did not leak, although it was difficult (perhaps impossible) to prevent some air being trapped inside the tube or released by the mercury. Boyle was right, however, to claim that his results were very similar to those previously obtained in experiments with barometers. If Boyle appeared to win in an English context it is because his most forceful opponent, Hobbes, was a more isolated, and thus less dangerous, figure than Descartes; Hobbes was fatally weakened by his reputation for atheism, while Descartes carefully placed his arguments within a wider context which was compatible with Christianity. It is important not to overstate the local and limited successes of Pascal and Boyle: belief in a vacuum, along with Copernicanism, ultimately triumphed only when Newton’s theory of universal gravitation (published in 1687) gave an account of how gravitational forces operated across empty space. Pascal and Boyle were then hailed as the discoverers of the vacuum which, it was now held, made up most of the universe. But in the 1660s the situation was quite different: in England, Henry Power, for example, continued to oppose Boyle’s claims on the basis of experimental evidence (and a sympathetic attitude to Cartesianism), just as Roberval had opposed Pascal’s.⁸¹

In 1661 Christiaan Huygens built his own air pump and began repeating the standard experiments. He tested the quality of his machine by introducing a water barometer to provide a sensitive measure of how much air (if any) remained in the experimental space. Huygens pumped away, air was removed, but the water level failed to fall. The tube remained full. The water Huygens was using had been purged of air in order to ensure it would not release air into the experiment; Huygens found that he could only get it to behave as expected if he introduced a bubble of air into it. When reports of this reached Boyle, he (quite naturally) dismissed them as nonsensical, but Huygens came to London and showed that the same results could be produced on Boyle’s apparatus. (The cause of this puzzling phenomenon is the tensile strength of water, without which trees could not grow above ten metres tall.) In the light of his results, Huygens abandoned his previous belief in a vacuum (despite the fact that the anomaly disappeared when a bubble of air was introduced into the water, he decided that some previously unknown substance was supporting the column of water), while Boyle decided to carry on as if nothing had happened.⁸²

What is crucial here is that the two conflicting results did not properly have equal status. The void-in-the-void experiment had been carried out time after time in barometers, with water, wine and mercury, and it had been performed with at least five different configurations of the apparatus, yet no result comparable to that produced by Huygens had ever been seen. Boyle therefore decided to see if he could produce anomalous suspension with mercury (something Huygens had not managed to do) by carefully purging it of air, as this would make it evident that there was something misleading about Huygens’ result. As he put it: ‘[T]he sustentation of tall Cylinders of Mercury in the Engine [i.e. the air pump] seem’d to me to have too little Analogy with all the Experiments yt hath been hitherto made about those of Torricellius.’⁸³

But even before performing the experiment with the air pump, Boyle succeeded in producing anomalous suspension of mercury (to a height of 52 inches) in the open air. At this point it was clear that the phenomenon had nothing to do with what was or was not present within a supposed vacuum. Huygens agreed: it took less than two years (a shorter period of time than it would now seem, given the difficulties of travel and communication in the seventeenth century) for him to accept that his anomalous result was irrelevant. One can reasonably say, then, that Huygens had been wrong to claim that his result was in some way ‘better’ than Boyle’s, and wrong to abandon his earlier beliefs in the light of it. Boyle’s result was the right result, and Huygens’ result was simply a deeply puzzling anomaly; we can say this with confidence now, but it was also apparent to intelligent observers at the time, and to Huygens himself as soon as Boyle demonstrated the anomalous suspension of tall cylinders of mercury, both in and out of the air pump.⁸⁴

The experimental method depends on independent replication, and one claim made by sociologists of science is that truly independent replication never takes place: in order to make a new experiment work, it is claimed, scientists always have to spend time in the company of scientists who have already performed it, picking up unwritten tricks of the trade. But Petit together with Pascal replicated the Torricellian experiment independently, and Valerio Magni in Warsaw either replicated it in 1647 or reinvented it. Others seem to have performed the experiment entirely independently, on the basis only of written descriptions – Henry Power, for example. The simple truth is that replication of the Torricellian experiment was unproblematic; it follows that the sociologists of science are wrong.xvixvi

If we look back over the history of experimentation in this way we can begin to understand the significance of what happens in the seventeenth century, symbolized by the small crowd which accompanied Périer to the top of the Puy-de-Dôme. Why were they there? Périer was surely glad to have witnesses, but they were there because they believed this was an opportunity to watch history being made. Their presence marks the beginning not of discovery itself but of a culture of discovery, a culture which was now shared by government-office holders and sophisticated clergymen. (Périer names two clergymen, two government-office holders and a doctor as accompanying him.) Moreover, publication ensured that there was a much larger number of ‘virtual onlookers’. As Walter Charleton put it, the little group of French experimenters ‘seemed unanimously to catch at the experiment [of Torricelli], as a welcome opportunity to challenge all the Wits of Europe to an aemulous [actuated by a spirit of rivalry] combat for the honour of perspicacity’.⁸⁵ And when Boyle coined the phrase experimentum crucis to honour the Puy-de-Dôme experiment he was marking the beginning of a new era in which philosophical disputes would be resolved by experimentation.

§ 7

The issue of replication is central to a topic of foremost importance for any understanding of the Scientific Revolution: the demise of alchemy.⁸⁶ Boyle and Newton devoted an enormous amount of effort to alchemical researches. Boyle seems to have spent much of his life trying to turn base metal into gold, although our knowledge of his activities is limited because most of the relevant papers were (as best we can tell) destroyed on the instructions of his first biographer, Thomas Birch.⁸⁷ Boyle believed he was on the verge of succeeding, so close indeed that he thought it prudent to campaign (successfully) for a change in the law, which condemned to death anyone making gold.⁸⁸

Like many alchemists, Boyle was convinced that the quest for the philosopher’s stone (which would turn base metal into gold) involved a spiritual element. He believed he had seen transmutation performed; he evidently thought it likely that the anonymous stranger who had performed it, in his presence and with his assistance, was an angel, no less.⁸⁹ He had been singled out for this special revelation. Such beliefs made Boyle the perfect target for the sophisticated con artist. One part of Boyle’s correspondence concerning alchemy survives by chance: it was written in French, a language unknown to Henry Miles, Birch’s assistant, who was tasked with sorting through Boyle’s papers and deciding which ones to throw out. From it, we learn that a Frenchman called Georges Pierre persuaded Boyle that he (Pierre) was the agent of the Patriarch of Antioch, the head of a society of alchemists that had members in Italy, Poland and China. To become a member Boyle had to hand over his own alchemical secrets, but also valuable gifts – telescopes, microscopes, clocks, luxurious fabrics, large sums of money. In return, Pierre reported on the manufacture of a homunculus in a glass vial. Pierre told a good story: one meeting of his secret society had, he assured Boyle, been disrupted by disgruntled employees who had blown up the castle in which the society was meeting. And Pierre went to great lengths: he planted stories about the Patriarch of Antioch in Dutch and French newspapers on the offchance that Boyle would come across them. In fact, when Pierre was supposed to be in Antioch, he was actually in Bayeux, having a jolly time with his mistress. And his wild stories had already acquired him the nickname in his home town of Caen of ‘honest Georges’.⁹⁰

How could Boyle, one of the key figures of the Scientific Revolution, be so totally convinced of the reality of alchemical transmutation? The answer is that alchemy was a self-fulfilling enterprise. Those who practised it were convinced that the philosopher’s stone had been successfully produced in the past. As George Starkey, who collaborated closely with Boyle in his alchemical researches, put it: ‘[T]he wise Philosophers with all their might have sought & found, & left the record of their search in writing, withall so veyling the maine secret that only an immediate hand of god must direct an Artist who by study shal seeke to atteyne the same.’⁹¹ Like Boyle, Starkey believed he had held the stone in his hand, and with it he claimed that he had been able to turn base metals into gold and silver – or at least into a sort of gold and a sort of silver, for the gold had proved unstable, and the silver, though very like silver, weighed too much.⁹²

Starkey sought to discover this and other lost secrets by careful study of the alchemical texts, which were written, as he acknowledged, in a deliberately impenetrable language. Since early in the seventeenth century the word ‘hermetical’ (meaning ‘in the tradition of the mythical author Hermes Trismegistus’, a supposed contemporary of Moses, to whom numerous works were attributed) had been given a new meaning: those experimenting with chemicals began to refer to containers being ‘hermetically closed’, in other words, airtight, the word ‘hermetical’ becoming a pun on the idea of impenetrability.⁹³ When Starkey failed to achieve the results he hoped for (bankrupting himself and reducing his wife and children to penury in the process) it never occurred to him that the texts were wrong;⁹⁴ he was convinced he had simply misinterpreted them or failed to follow their instructions with sufficient precision. Thus alchemists had in principle procedures for verification (‘trial by fire’), although verification was constantly deferred. They had no procedure for falsification.

Starkey referred to the adage ‘Hear the other side’, a fundamental principle of natural justice, as ‘hateful’.⁹⁵ The other side he refused to hear were those who dismissed alchemy as a deception, a delusion, a fantasy. Yet among scholastic philosophers these were the majority, from Aquinas and Albertus Magnus on. Alchemy had long been mocked by sceptics – by Reginald Scot in the Discovery of Witchcraft (1584) and by Ben Jonson in The Alchemist (1610) to take but two examples among many. Belief could be sustained only if a peculiar authority was granted to obscure books, or better still ancient manuscripts newly discovered in locked chests. ‘Alchemy was inalienably a textual as well as an experimental science,’ writes Brian Vickers, and in paintings the alchemist was always shown surrounded by books or manuscripts as well as by the impedimenta of his laboratory.⁹⁶ Missing from the paintings, though, is the most important moment of all, the moment when one person was persuaded to put their trust in another. Thus Boyle left at his death ‘a kind of Hermetic legacy to the studious disciples of that art’ (a legacy which does not survive, and was presumably destroyed). Included were many alchemical recipes he had not tried but which he was sure were efficacious, for they had been ‘(though not without much difficulty) obtained, by exchange or otherwise, from those that affirm[ed] they knew them to be real, and were themselves competent judges, as being some of them disciples of true adepts, or otherwise admitted to their acquaintance and conversation’.⁹⁷ Difficulty was itself the guarantor of authenticity; in the absence of anyone who could reliably be identified as a true adept (i.e. someone capable of producing the philosopher’s stone) a mere claim to acquaintance and conversation was sufficient to authenticate an incomprehensible text as carrying hidden meaning. Boyle believed because he wanted to believe.

In the recent literature there has been an extended effort to present alchemy (or, as its historians now prefer to call it, ‘chymistry’) as the first experimental science; out of alchemy, we are told, came modern chemistry.⁹⁸ In showing that many alchemical recipes can indeed be performed in a modern laboratory, these scholars have rendered apparently incomprehensible texts meaningful and reinstated alchemy as a laboratory science. But if this argument is pressed too far it becomes difficult to explain why, in the eighteenth century, modern chemistry established itself not as a continuation of but as a refutation of alchemy. Why did Birch destroy Boyle’s alchemical papers, not celebrate them?

There has been little written on the end of alchemy, yet an activity which was respectable in the eyes of Boyle and Newton had become entirely disreputable by the 1720s.⁹⁹ John C. Powers has argued that this was the result of a series of ‘rhetorical’ moves by chemists in the Académie des Sciences, such as Nicolas Lémery (1645–1715), who adopted many of the experimental findings of the alchemists, and also many of their attacks on those who brought their art into discredit by telling tall tales; at the same time they dismissed as ridiculous the quest for the philosopher’s stone. The implication is that deep down they were alchemists, they were just not prepared to admit it. Powers does not consider taking the eighteenth-century chemists at their word. The advocates of the new chemistry insisted that they had no time for texts whose meaning was impenetrable. They protested that ‘that sect of chemists [i.e. the alchemists] … writes so obscurely that in order to understand them one must have the gift of divination.’¹⁰⁰ They were, they insisted, interested only in chemical processes which they could reproduce in their own laboratories and could then have certified by their colleagues. ‘Each memoir,’ produced by the advocates of the new chemistry, writes Powers, ‘presented a limited investigation into a specific question or set of questions, and the chemist relied solely on the account of his experiments to persuade his audience [to] accept his conclusions.’¹⁰¹ Powers describes these as ‘purported’ experiments, but of course they were real.

What made it possible to consign alchemy to the dustbin of history was a new understanding of what chemists were trying to do. For the alchemists, including Boyle and Newton, the fundamental enterprise was one of transmuting one substance into another. But in 1718 Étienne François Geoffroy, the son of a pharmacist and the holder of the chair in chemistry at the Jardin des Plantes in Paris, an institution established for the training of pharmacists, published a ‘Table of Different Relations Observed in Chemistry’. Geoffroy’s table lists what he calls ‘the principal materials with which one usually works in chemistry’ (a total of twenty-four), but he left out all sorts of substances that chemists frequently worked with. The principle of selection is revolutionary: the materials he lists combine with each other to form new stable compounds – but each of these compounds can be broken down, if the right chemical procedures are followed, to release their original components. Geoffroy’s twenty-four substances thus survive even when they are combined with other substances: they are not transmuted when they enter into such combinations. Geoffroy was a long way from having a modern theory of elements of the sort propounded by Lavoisier towards the end of the century, but he did have a research programme which had escaped entirely from the concept of transmutation. It is thus Geoffroy, not (as is often claimed) Boyle, who marks the beginning of modern chemistry.¹⁰²

Geoffroy’s work appeared in a context where chemists were already trying to escape from alchemical thinking. What killed alchemy was not experimentation (Starkey, Boyle and Newton were indefatigable in their pursuit of experimental knowledge), nor the development of learned networks devoted to new knowledge (alchemists were very effective at seeking each other out and worming information out of each other, always on the basis of exchanging one secret for another), nor even Geoffroy’s recognition that chemical combination did not imply transmutation. What killed alchemy was the insistence that experiments must be openly reported in publications which presented a clear account of what had happened, and they must then be replicated, preferably before independent witnesses. The alchemists had pursued a secret learning, convinced that only a few were fit to have knowledge of divine secrets and that the social order would collapse if gold ceased to be in short supply. Some parts of that learning could be taken over by the advocates of the new chemistry, but much of it had to be abandoned as incomprehensible and unreproducible. Esoteric knowledge was replaced by a new form of knowledge which depended both on publication and on public or semi-public performance. A closed society was replaced by an open one.xviixvii

If in thinking about alchemy we keep our eyes on individuals like Boyle, we are in danger of missing the role of institutions, both formal, such as the Royal Society and the Académie des Sciences and informal, such as Mersenne’s circle. Many of the founding members of the Royal Society – Digby and Oldenburg, for example, as well as Boyle – were preoccupied with alchemy. But alchemical transmutations were never discussed at the meetings of the Royal Society, and only one brief publication by Boyle in the Transactions dealt with alchemical matters; it served effectively as an advertisement, announcing his interest in the hope that others would contact him.¹⁰³ Everyone (except perhaps Boyle) was clear that the principles on which the Royal Society was based – the free exchange of information, the replication of experiments, the publication of results, the confirmation of ‘facts’ – were at odds with the principles of the alchemists. Boyle and Newton were both alchemists and participants in the new scientific community, but for the most part they were perfectly clear that the two sides of their lives were separate, just as Pascal was clear that his religious life, which was intense and demanding, was separate from his scientific life. Boyle, it is true, wanted to bring alchemy a little into the public eye, if only in order to make it easier for alchemists to identify each other – Newton immediately told him off, recommending ‘high silence’. Boyle, he complained, was ‘in my opinion too open & too desirous of fame’.¹⁰⁴

Pascal, as we have seen, maintained that the fundamental difference between science and religion was that in science there were no truths that could not be questioned, while religion depended on accepting certain truths as being beyond question. For the alchemists, the reality of the philosopher’s stone was beyond question; within a generation their appeal to authority, to ancient texts and secret manuscripts, seemed hopelessly misplaced. Alchemy was never a science, and there was no room for it to survive among those who had fully accepted the mentality of the new sciences. For they had something the alchemists did not: a critical community prepared to take nothing on trust. Alchemy and chemistry were both experimental disciplines, but the alchemist and the chemist had different forms of life and belonged to different types of community.xviiixviii An important consequence follows from this argument: we should not really expect to find reliable science before scientific communities began to take shape in the 1640s. And this seems right. Had Galileo belonged to a functioning scientific community, to take just one example, he would have been firmly discouraged from placing his theory of the tides at the centre of his defence of Copernicanism.¹⁰⁵

Consequently, we do not have to wait until the publication of Geoffroy’s table in 1718 to hear the death knell of alchemy. According to the new historians of chymistry, alchemy and chemistry were a single undifferentiated discipline until the publication of the third edition of Nicolas Lémery’s textbook in 1679, when distinctions between them began to be drawn;¹⁰⁶ by the 1720s the two had been effectively separated. Here, though, is the Plus ultra of Joseph Glanvill, published in 1668. It contains fulsome praise of Boyle as someone whom pagans would have worshipped as a god, but its approach to alchemy/chymistry clearly foreshadows that of the eighteenth century:

I confess, Sir, that among the Aegyptians and Arabians, the Paracelsians, and some other Moderns, Chymistry was very phantastick, unintelligible, and delusive; and the boasts, vanity, and canting of those Spagyrists [alchemists], brought scandal upon the Art, and exposed it to suspicion and contempt: but its late Cultivatours, and particularly the ROYAL SOCIETY, have refin’d it from its dross, and made it honest, sober, and intelligible, an excellent Interpreter to Philosophy, and help to common Life. For they have laid aside the Chrysopoietick [gold-making], the delusory Designs and vain Transmutations, the Rosie-crucian Vapours, Magical Charms, and superstitious Suggestions, and form’d it into an Instrument to know the depths and efficacies of Nature.¹⁰⁷

Glanvill would presumably have been shocked to learn that Boyle and Newton did not share his views, but it was he, not they, who had grasped the relationship between alchemy and the new science. The Lexicon technicum of 1704 expressed the rapidly emerging consensus:

ALCHYMIST, is one that studies Alchymy; that is, the Sublimer Part of Chymistry which teaches the Transmutation of Metals and the Philosopher’s Stone; according to the Cant of the Adeptists, who amuse the Ignorant and Unthinking with hard Words and Non-sense: For were it not for the Arabick Particle Al, which they will needs have to be of wonderful vertue here, the word would signifie no more than Chymistry. Whose Derivation see under that Word. This Study of Alchymy hath been rightly defined to be, Ars sine Arte, cuius principium est mentire, medium laborare, & finis mendicare: That is, an Art without an Art, which begins with Lying, is continued with Toil and Labour, and at last ends in Beggery.¹⁰⁸

A family of alchemists at work, an engraving by Philip Galle, after a painting by Pieter Bruegel the Elder, published by Hieronymus Cock, c.1558.

The demise of alchemy provides further evidence, if further evidence were needed, that what marks out modern science is not the conduct of experiments (alchemists conducted plenty of experiments), but the formation of a critical community capable of assessing discoveries and replicating results. Alchemy, as a clandestine enterprise, could never develop a community of the right sort. Popper was right to think that science can flourish only in an open society.¹⁰⁹