2.1Numbers as a substitute for trust
Porter calls quantification a technology of distance: rules for counting and calculating let communication reach beyond a community of trust. He names the pursuit of such rules mechanical objectivity, following rules rather than judgment. His most pointed claim concerns who pursues it: not secure elites, but weak or contested experts and bureaucracies under outside pressure, for whom rules stand in for personal trust (Porter 1995).
Rusnock's history of vital accounts gives this argument an eighteenth-century setting. Physicians and others in England and France, she shows, used the bills of mortality and new tallies of their own to settle medical and political questions. The best-known was whether inoculation against smallpox was worth its risk, which they tried to settle by comparing the deaths from natural smallpox with those from the inoculated kind. The trust such accounts won, on her account, depended on how the numbers were gathered and on who vouched for them (Rusnock 2002).
The two accounts are close but not identical. Porter stresses rules that make the person unnecessary; Rusnock stresses gathering and vouching, which keep the person in view. A count that persuades by its rules should state its definitions so that anyone could apply them; one that persuades by vouching should say who gathered the numbers and why they can be believed.
2.2Numbers as the making of states
Desrosières's history starts from a different place. To count anything, one must first decide which cases are the same kind of thing. That decision, a convention of equivalence, is at once a technical choice and a political one. Once it is established, it holds, and the category becomes a thing that can be acted upon. His history treats statistics as bound up with the making of states (Desrosières 1998).
Scott's argument concerns what states do with categories once they have them. States make society legible by simplifying it, through standard names, maps, measures and registers. Scott's further claim is that high-modernist schemes which impose such simplifications and override local, practical knowledge, which he calls mētis, have failed, sometimes disastrously (Scott 1998).
Hacking's history sits between the two. It traces how, from about 1820 to 1840 especially, an avalanche of printed numbers (official counts of births, deaths, diseases, crimes and suicides) led to the idea that society was governed by statistical laws, and how the categories used for counting shaped what was counted. Determinism eroded as chance came to be seen as lawful (Hacking 1990).
2.3The debate laid side by side
The disagreement can be put as two predictions about a mortality count read closely. On Porter's side, with Rusnock, its definitions are instruments of travel, stated so that strangers can check them; its authority rests on the rules and on the visible way the numbers were gathered; and the counter writes as someone whose personal standing is not enough. On the side of Desrosières and Scott, its definitions are acts of classification that decide which deaths are alike and so make them a thing a government can act upon, and, in Scott's version, override the knowledge of those closest to the matter. Hacking adds that the categories shape what is counted at all.
These predictions are not exclusive. A definition can be both a rule for strangers and a convention that serves a state. What the two documents can show is which of these functions the authors themselves claim, and whether the knowledge of those closest to the deaths is overridden, used or corrected.
2.4The plague, the searchers and the town under watch
Graunt's bills belonged to a system of plague administration. Paul Slack's social history of plague in England examines the plague orders issued from the 1570s, which shut up infected households, appointed searchers and watchmen and kept bills of mortality. These measures fell hardest on the poor, provoked resistance, and divided magistrates, clergy and physicians over whether isolation was just (Slack 1985). Michel Foucault, in the chapter of Discipline and Punish called Panopticism, describes the regulations a late-seventeenth-century town applied when the plague came: the town partitioned, each household shut in and inspected daily, every death and illness registered and passed up a hierarchy. He treats this plague model of segmentation and constant surveillance as an image of disciplinary power (Foucault 1977).
Both accounts place the count inside an apparatus of control, and Graunt's text can be asked whether the bills he read flowed only upward.
2.5Nightingale: persuasion or discovery
The 1858 report has its own debate. I. Bernard Cohen presents Nightingale as a statistician who used statistics as a tool of reform, presenting the army's mortality in graphical forms, among them the polar-area diagram, to persuade officials that most deaths in the Crimean War came from preventable disease (Cohen 1984). Hugh Small argues that the very high death rate at the Barrack Hospital at Scutari in the winter of 1854 to 1855 came chiefly from its sanitary defects (drains, ventilation, overcrowding), which the Sanitary Commission sent out in the spring of 1855 began to remedy, and that Nightingale came to understand this fully only after the war, when she analysed the army's mortality statistics with William Farr; from then on she campaigned for sanitary reform (Small 1998).
Side by side, the two accounts give the count a different place in the story. For Cohen the statistics are an instrument of persuasion aimed at officials. For Small the analysis of the statistics is also where understanding came from, and it came after the events it explained. The debate is relevant here because a count that persuades and a count that teaches its own maker must settle their denominators for different reasons.
2.6Counting while it happens
A last body of work concerns the biases of counting deaths during an outbreak. Marc Lipsitch and colleagues set out why estimates of the share of cases who die, made while an outbreak is under way, are biased: deaths lag behind cases, so early ratios understate the share; mild cases are missed, so ratios overstate it; and case definitions, testing and reporting differ between places and over time. Comparisons across places and periods, they argue, need these biases taken into account (Lipsitch and others 2015). The paper concerns modern outbreaks and is used here only as a vocabulary for the kinds of distortion a count can carry.