EN
Radiocarbon dating is a typical example of physical measurement applied in archaeology. As in all measurements, the important thing is the random factor of measurement error. The error is determined by Gaussian probability distribution applied to the actual age of a sample (which while determined, remains unknown). This situation, which is typical of the natural sciences, physics and technology, is easy and even banal compared to what archaeologists had to face once calibration was introduced in the radiocarbon method. Presently, the complexity of the probability distribution of the actual age of the sample (already the ordinary calendar age) matches that of the form of the so-called calibration curve. It is not by accident that advanced methods of strict taking into account of a priori uncertain information, based on Bayes' theorem, have found application in archaeology. Inferences made with the benefit of high mathematical discipline make it possible to interpret as fully as possible the information contained in a calibrated date. Yet the information provided in a date hardly warrants the complicated form of the probability distribution of a calibrated age, so often cited in extenso in archaeological research. At this point matters take on a difficult turn. Computer programs for calibrating radiocarbon dates present the range of estimation in which the actual age of the sample falls with the set probability. There are two probabilities in use: 0.68 and 0.95. Archaeologists are recommended to make use of the effect of the dating in the simplest form possible: as a range corresponding to the 0.95 probability, with the assumption that it is a continuous range uninterrupted by sections of slightly lower probability values. Assuming ideal sampling and laboratory measurement conditions, it will still give an average of one date in twenty that will fall outside the given range (which, of course, we shall never be aware of).