On Inexact Dates
The advice we all absorb during graduate study is to give dates in our published work in such a way as to minimize the chance that they will later be found wrong. This, unfortunately, leads to routine vagueness in scholarly writing, and to the assigning of uncertain dates to centuries (or millennia) rather than years.
But century dating has implications that may not reflect the researcher's best opinion. Such a date as "03c" (3rd century BC; for the prefixed-zero convention, see our Dates page) says this to a reader: "somewhere in the span 0200/0299, but equally likely to be any of those years." The situation where we are certain that a work may be from 0299 but cannot possibly be from 0300 (and similarly at the other end of the span), but have no clue as to where within the span 0200/0299 it might be, is surely rare. The shape of our uncertainty is usually somewhat different. That sort of uncertainty is not well reported by the "century" convention.
Scholarly progress as a whole has higher priorities than the minimizing of personal embarrassment. It does not want vagueness; it wants to know people's best opinion, so as to subject it to scrutiny and improve it as rapidly as possible. A date like "c0300" might be anything in the two adjacent centuries. It will draw no comment. A date like "c0317" is much more likely to prompt such a comment as "but it cannot be later than Mencius's visit to Ngwei in 0320," with resulting improvement in the precision of the guess. Progress needs comment of this sort. The question for the methodologically conscious student is then this: How can our best sense of a date be most efficiently be conveyed to readers and colleagues for their criticism?
The responsible answer is that one should indicate as precisely as possible the elements of both certainty and uncertainty. But conveying that information might easily require a paragraph, which is impossibly cumbersone. Over the centuries there have evolved some shorthand conventions for the various possibilities. Here are our recommendations for how to notate those possibilities.
- Date: 0254. When the evidence suggests a precise year.
- Circa Date: c0254. "Most likely 0254; maybe a little earlier or later"
- Center Date: (c0254 ± 10). "0254, within ten years either way" (± means "plus or minus")
- This gives our best guess as to the range of the uncertainty
- There is no "standard deviation" implication; compare Radiocarbon Dates below
- Straddled Year: (0255/0254). Or, "During the first Chu conquest of Lu"
- That conquest began late in 0255 and was completed in early 0254
- The second and final conquest phase was carried out in 0249, whence:
- Range: (0255/0249). "Somewhere during the two-phase Chu conquest of Lu"
- Another example: 0319/0301, "in the reign of Chi Sywaen-wang"
- The error-minimizing guess is the center date, c0310 ± 9. Compare:
- Asymmetric Range: 0319/0313/0301. "Probably 0313 within Sywaen-wang's reign"
- 0313 is the year after the Yen episode, which was a disaster for Chi
- Span: 0549-0479. The entire period, here the life span of Confucius
- Or with refinements, such as c0549-0479 or 0552/0549-0479
Our recommended device for indicating partial certainty, with only one open endpoint, is a hyphen placed before or after the date (note that this convention is only possible if one is not using a minus sign to indicate "BC"). There is a well established terminology for single-ended dating, as follows:
- Terminus ad quem: -0249. In some year up to and including 0249
- The year given is the latest possible one
- Equivalent to: Terminus post quem non ("not later than"). See next group
- Terminus a quo: 0255-. In 0255 or some later year
- The year given is the earliest possible one
- Equivalent to: Terminus ante quem non ("not earlier than"). See next group
For noninclusive endpoints, we might have instead
- Terminus ante quem: -//0249. Up to but not including 0249
- Terminus post quem: 0249//-. Later than but not including 0249
where the two slashes // make a wall between the fixed date 0240 and the zone of uncertainty. These four devices exhaust the logical possibilities.
Radiocarbon dates are properly given in the form: 0376 ± 80. This has a statistical meaning. It is a center date plus a one standard deviation range. It may be paraphrased thus: "The most likely date is 0376; there is a 68% chance that the actual date is within 80 years of that." The uncertainty range has nothing to do with antiquity; it refers to the probable outcome of repeated tests on the same material by the same lab.
NB: Radiocarbon dates are sometimes supplied from the lab as BP or "Before Present;" the "present" being in fact the base year 1950. See Bowman: Radiocarbon Dating, California 1990. These lab dates should be converted to historical dates before being used in historical writing.
There are some genuine radiocarbon uncertainties. A given radiocarbon lab result may equally well refer to any of three widely-separated historical years, due to variation both up and down in the carbon-14 level in the environment. This gives rise to the "calibration" problem; see Bowman p46f. It has nothing to do with uncertainty about the lab analysis per se. It is related to the conditions obtaining in antiquity.
Unfortunately, the Warring States period falls within one of those zones where multiple year possibilities for a given radiocarbon reading exist, and radiocarbon dates should thus be used with special care.
7 April 2000 / Contact The Project / Exit to Reference Page