The danger that comes readily to mind can seem the danger most likely to arrive. An explosion supplies a scene; a slowly accumulating injury supplies a statistic. The scene has faces, a place, a moment at which an ordinary day became irretrievable. The statistic must be learned and remembered without that assistance. The earlier plate on availability followed this substitution from the frequency of an event to the ease of recalling it. Here the substitution enters a public decision: which hazards deserve attention, expenditure and restraint? There are two questions inside that decision. One concerns how often harm occurs. The other concerns what makes the harm objectionable. Vividness can mislead the first judgment. A greater fear of catastrophe can also express something the count alone has left out. To understand the departure, we must first establish which judgment the person is making.
Sarah Lichtenstein, Paul Slovic and their colleagues found both competence and substantial error when they asked people to judge the frequency of lethal events. Their 1978 study distinguished a broad compression of frequencies—rarer causes tended to be overestimated, commoner causes underestimated—from additional differences among particular causes. Sensational causes often stood above the relation between judged and recorded frequency; quieter causes stood below it. The distinction matters. Calling every overestimate an effect of drama would confuse the general compression with the particular event's prominence. The authors considered several explanations, including anchors, uneven exposure to events and the ease with which examples are stored or retrieved. Their historical experiments establish patterns in the judgments they studied. They do not establish a permanent ranking of everyone's fears, or identify news coverage as the sole cause of every mistaken estimate.1
A simple reporting rule can reverse the apparent order even when every reported event is true. Suppose, illustratively, that two kinds of harmful incident occur in the same population during the same year. There are ten incidents of the first kind and a hundred of the second. All ten of the first kind are reported, but only one of the second is. Someone counting the reports encounters a ratio of ten to one where the underlying ratio is one to ten. Nothing has been invented; selection has done the work. If instead the same tenth of each category were reported, one report would represent the first kind and ten the second, preserving the order. The example isolates a problem in the evidence reaching memory. It does not show that everyone mistakes reporting for frequency. It shows why a truthful collection of vivid instances can still be an unreliable basis for a judgment about how often harm occurs.2
The word risk, however, can ask for more than a frequency. In his 1987 account of risk perception, Slovic described research in which people rated hazards along dimensions such as control, familiarity, catastrophic potential and delayed consequences. Two recurring combinations were dread and the unknown: the prospect of a fatal, uncontrollable catastrophe, and the prospect of an unfamiliar harm whose operation or eventual effects could not readily be observed. Lay judgments of risk could differ from those same people's estimates of annual deaths, because the judgments included these other characteristics. The difference is therefore not always a miscalculation of mortality. It can be a disagreement about which properties of a hazard count. Slovic expressly denied that the resulting map was a universal arrangement of every danger. It was a way of studying the structure of concern, with limits to its reach and predictive authority.3
Equal expected losses need not describe equal hazards. Consider fifty people, each exposed for the same period. In one illustrative arrangement, a common accident kills all fifty with a probability of two percent and otherwise kills no one. In another, each person independently faces a two-percent probability of death. Each arrangement has one expected death, and each person has the same marginal probability of dying. Yet one puts all the deaths into a single catastrophe; the other spreads the possibilities across independent events. A calculation that records only the mean has deliberately suppressed this difference. Whether the concentration warrants greater precautions depends on the consequences and the evaluative rule, including any additional harm to a community when many lives are lost together. Neither a larger emotional response nor an identical mean settles that question. The arithmetic identifies what is equal; judgment must explain which remaining differences matter.4
A dramatic event can also supply new information. Slovic treated accidents as signals: a small visible failure may disclose something about a larger system, its safeguards or the institutions responsible for it. If the accident reveals a failure mode previously thought impossible, revising a risk estimate is different from counting the same report repeatedly. Conversely, repetition without new evidence need not justify a larger estimate. This distinction preserves the informational value of attention while making it answerable to a mechanism. What has been learned about the process producing the harm? Does the event belong to the same population and conditions as the earlier estimate? Which consequences were already counted, and which have become newly apparent? Asking those questions neither dismisses public concern nor treats concern as its own proof. It directs the vivid image back toward the hazard it may illuminate.5
The vivid can make a rare event seem common, and it can make an overlooked feature of a common event impossible to ignore. A sound reckoning has to distinguish these achievements. For a frequency claim it asks for a population, an exposure, a period and a method of counting. For a judgment about acceptable risk it also asks who bears the danger, who receives the benefit and what consequences the chosen measure excludes. Those inquiries can change the policy without changing the number of deaths, or change the estimated number without settling the policy. The image is evidence of what has seized the mind; its relation to the world remains to be established. That is why the final plate must return to the standard itself: before calling a fear excessive, or a choice a departure, we must say what the baseline measures.