by John Dunn and Kim Kirsner, available here.
The authors analysed the statements of the survivors of a naval battle to infer the potential location of the wrecks. Although the statements contradict each other, the authors assumed that they were essentially corrupted versions of genuine location information, coming for example from the ship’s captain and having mutated through word-of-mouth or misremembering. They used some simple quantitative models – and also a lot of informed guesswork – to figure out what the source statements were. The approach is quite reminiscent of work done on reconstructing the history of languages using evolutionary models, à la Atkinson & Gray. Well worth reading.