Time isn't kind
(I didn't have enough time to fish out my intended title, which was to be from the riddle match between Gollum and Bilbo in The Hobbit. Time, time!)
Becoming a fossil requires hard work and a lot of luck. It's not enough to die in a nice mushy riverbank, although that's a start. Your remains could get obliterated, washed away, or dispersed, before they even get properly mineralized. And, eons later, you could get ground up by a Chinese apothecary, sold on Ebay, or waylaid in any number of other ways instead of being dug up, polished, argued over, and shelved.
It's very inefficient, but how inefficient? In fact all the oddball events in a fossilizable entity's natural history tend to average out. So paleontologists have had some luck in modeling the "population" of fossil relics from a given population over time. The equations make certain idealizations including a constant environment and a closed, homogeneous population, reproducing exponentially, and subject to no more than normal post-fossilization vagaries.
In last week's Science (subscription link; free link at the BBC is here ) these equations have been ported over to describe what seems to be a completely different phenomenon: the survival of medieval parchment manuscripts through the looting and pillaging of the ages. The gauging of how many are left behind has been the province of anecdote and wild guesses, but these fossil equations seem to describe what to expect rather well. The reason it works for manuscripts, as summarized by Sharon Gilman and Florence Glaze, is that, prior to the printing press, the numbers of technical manuscripts behaved like a biological population: they were reproduced exponentially (as copies were made of copies of copies), up to a saturation point, and destroyed with some probability. Death and destruction, although seemingly arbitratry, actually average out over time. Using the equations derived for predicting fossil survival, the author predicts that a very big fraction of technical works (he suggests 2 in 7) have survived since the parchment era.
In their fairly snarky synopsis of the work ("If he had teamed up with a historian in the first place, he could have written a much better essay"), Gilman and Gaze point out some limitations to the modeling regarding the idealizations necessary for the math. Manuscripts do not provide a homogeneous population (every text is different, and therefore copied and disseminated differently from the others); the population is not closed; and it hasn't been long enough for the "noise" in the processes of destruction to have averaged out from text to text. Still, I think it's a fascinating first stab, and the match between model and reality for Bede's manuscripts looks pretty impressive. Now, back to my parchments...