Six passages, 24 questions

A naval surgeon’s trial of 1747, in which twelve scorbutic sailors were assigned in pairs to six different treatments and only the pair given citrus recovered, is commonly presented as a decisive experiment that an obstinate profession ignored for half a century. The delay is real. The explanation is not adequate.

The result was unambiguous only in retrospect. Citrus was one of six candidate remedies, and the surgeon did not conclude that it acted specifically; he attributed its effect to an acidity it shared with several other substances, and recommended a preparation — a concentrated syrup, boiled down for shipboard storage — that destroyed the property actually responsible. Ships supplied with the syrup saw no benefit, and this was taken as evidence against citrus rather than against the preparation.

The institutional obstacle was of a different kind again. Scurvy appeared on long voyages and not on short ones, and the administration’s difficulty was not identifying a remedy but procuring, storing and distributing a perishable commodity across a fleet at sea for years. When lemon juice was eventually issued as a matter of course, the change followed not a fresh demonstration of efficacy but a contract with a supplier able to deliver at scale.

None of this makes the profession’s conduct admirable. It does suggest that the standard account misidentifies what was missing. A demonstration that a substance cures a disease is one input to a decision that also requires a theory of why it works, a preparation that preserves the active property, and a supply chain capable of delivering it. The half-century is better read as the time taken to assemble the other three.

The steel produced in south India and Sri Lanka from about the third century, and traded across the Indian Ocean under a name that reached Europe as wootz, was made by a process whose details were never recorded and have been reconstructed from crucibles, slag and the finished metal. That reconstruction has proceeded in two directions, and they disagree about what the process was for.

The first holds that the crucible method was a means of producing steel of controlled carbon content, which the bloomery processes used elsewhere could not reliably achieve. On this account the sealed crucible allowed the smith to fix the proportion of carbon by charging the vessel with measured quantities of iron and carbonaceous material, and the value of the product lay in its consistency.

The second holds that consistency is not what the process delivered. Analyses of surviving blades show carbon contents varying between one and one and a half percent within single objects, a range wide enough to alter the metal’s behaviour under the hammer. What the crucible did deliver, on this account, was a melt: iron in the liquid state, which the bloomery never achieved, and which allows slag to separate and impurities to segregate in ways that solid-state processes do not.

The disagreement matters beyond metallurgy, because the two accounts imply different things about what the smiths knew. Controlling carbon content requires a quantitative conception of the charge. Achieving a melt requires only sustaining a temperature, and the temperature is signalled by the appearance of the crucible. The material evidence at present favours the second, though by a margin narrow enough that a single well-preserved workshop could reverse it.

Mortality in western Europe fell substantially between 1850 and 1940, and the fall preceded the therapies now credited with controlling the diseases responsible. Antibiotics arrived after most of the decline in tuberculosis deaths had already occurred, and the same holds, with variations, for several other infections.

The observation has been used to argue that clinical medicine contributed little, and that improvements in nutrition and living standards did the work. The argument has force against a naive account in which each therapy produces a corresponding fall, and it has been influential in policy, where it is cited in support of directing resources away from clinical services.

It is weaker than its influence suggests. Nutrition and living standards are not a single variable, and the components that improved did not improve together: real wages, housing density, water supply, sewerage and the adulteration of food changed over the period on different schedules and in different places. Where those schedules can be separated — and municipal records sometimes permit it — the timing of mortality decline in a given city tracks water and sewerage more closely than it tracks wages. That is a public health intervention rather than a rise in living standards, and it is a deliberate one.

The distinction is not pedantic. An account in which mortality fell because societies grew richer implies that health follows growth. An account in which it fell because societies built sewers implies that health follows a decision about what to build, and that the decision is available at different points on the income curve. The historical evidence does not settle which framing is correct in general. It shows that the second is available, which the argument in its influential form obscured.

Earthquake prediction, in the sense of specifying the time, place and magnitude of a future event within useful bounds, has not been achieved, and the reasons are worth separating from the disappointment.

The programme rested on an analogy with weather. Meteorological forecasting improved as observation density rose and models grew more detailed, and it was expected that seismology would follow once instruments were dense enough. Precursors were sought: changes in ground water, in radon emission, in electrical resistivity, in the rate of small events. Several were reported before individual earthquakes. None has survived testing against the full record, in which the same signals appear frequently with no earthquake following.

The analogy may be the error. A storm is a large structure whose state can be sampled; an earthquake originates at a point on a fault at depth, where nothing can be measured directly, and whether a small rupture arrests after a metre or propagates for a hundred kilometres may not be determined by any condition distinguishable beforehand. If rupture size is settled by the details of the fault surface as the rupture encounters them, then no quantity of prior observation constrains it, and the failure of prediction is a property of the phenomenon rather than of the instruments.

This is not established, and it has the shape of a claim that would be difficult to falsify. What has changed is the direction of effort: from predicting individual events to estimating the probability of ground motion at a site over decades — which does not require knowing when a rupture will begin, and which is what a building code needs.

The attribution of a manuscript to a particular scribe rests on the assumption that handwriting is individual and stable, and palaeography has been practised for two centuries on that basis. Both halves of the assumption have been questioned, and they fail in different directions.

Stability fails within a working life. Scribes trained in one style and worked in another, and the hand of a copyist at twenty and at fifty may differ more than the hands of two contemporaries trained together. Where a dated sequence by a single scribe survives — which is rare — the drift is visible and substantial.

Individuality fails the other way. Institutional scriptoria trained their copyists to a house style, and the training succeeded well enough that the products resist separation. A study of one Cistercian house identified twelve hands in a corpus that the abbey’s own records indicate was produced by more than thirty men. The purpose of the training was to make the output uniform; where it worked, the palaeographer’s method is defeated by the scriptorium’s success.

The response has been to shift attention from letter forms to features less subject to training: the ruling pattern, the treatment of abbreviations, the handling of line-end division, the errors a copyist makes and repeats. These are more promising, though they carry the same difficulty in weaker form, since a scriptorium that trained letter forms also trained conventions. What the shift has made clear is that attribution to a named individual is rarely available, and that attribution to a workshop, a decade and a region is usually what the evidence will support.

Domestication of the founder cereals was long described as a rapid event: within a few centuries, wild populations gave way to crops carrying the traits — non-shattering ears, larger grains, uniform ripening — that make harvesting practicable. The rapidity was inferred from the strength of the selection that cultivation implies. A farmer harvesting with a sickle collects disproportionately from plants whose ears remain intact, and the arithmetic suggested that the non-shattering type should come to dominate within twenty to two hundred generations.

Archaeobotanical sequences do not show this. At sites where the transition can be followed through stratified deposits, the proportion of non-shattering ears rises over two to three millennia — an order of magnitude slower than the models predict. The discrepancy has proved robust to the addition of further sites.

Explanations have converged on the harvesting itself. Sickle harvesting of ripe stands is one method among several, and the others — beating ears into a basket, cutting green and ripening under cover, uprooting whole plants — impose weak selection or none. If cultivation was for most of its early history a supplement to gathering rather than a replacement for it, then the sickle was used intermittently and on part of the stand, and the effective selection was a small fraction of what the models assumed.

This reframes the question. The slow transition is not a puzzle about plant genetics but about what people were doing, and it suggests that the period usually described as the beginning of agriculture was a long interval in which cultivation and gathering were practised together without either displacing the other. That interval is longer than the agricultural revolution it was supposed to introduce.