The meandering nature of historiographical reconstruction has occurred to anyone who has worked in that field at all. One starts out at a general topic, such as the sources of a historical painting or the origins of Mendelian experimentation, and finds oneself, a couple of units of effort later, in the middle of contemplating the household receipts and textual editions (not to mention X-rays of the painting under consideration) or computing the total number of pea plants that the Moravian monetary garden (plus glass house) might have held or the names and origin of Mendel's four assistants (mostly lay brothers).
How did one get there?
Perhaps the most straightforward suggestion is afforded by David Lewis and his observation on the
asymmetry of overdetermination of actual worlds (1973). The observation here is the plurality of traces that causal events leave and the point is to allow the philosopher by comparison to establish the causal relationship with respect to the effects.
At issue is not the correctness of Lewis' observation per se but the proposal that actions in the real world that are causal leave a plurality of traces that extend into the historical record in a bewildering variety of ways. (The image of ripples on a pond is usually exercised at this point in the discussion.) All the same, time is the great destroyer and as the spans increase and other ripples chase across the pond, the traces become effaced. The artist's studio has disappeared, and so have his tools, materials and assistants; the Mendelian fields and glasshouse, should they have survived at all, were preserved only due to the conservative nature of the institution where he experimented. The historian is thus faced with "temporal cluster bombing" that works counter to the very tracing of the effect chains the historian wishes to undertake. At the same time, if the event was "big" enough (the extinction of the dinosaurs by meteorite comes to mind) the large number of traces prevents the erasure of every single detail---too much shocked quartz, too much iridium, too little volcanic activity in the Deccan traps.
My mother still remembers where she was and what she was doing the day John F. Kennedy was assassinated in Dallas, even though she was then an Austrian teen. It would be difficult to erase all traces, to spill damnatio memoriae on all the causal lines, that reverberated through the global village of modern media when that US president died. Diaries, letters, mementos, not to mention an avalanche of conspiracy theories all take their origin in that event.
It is this situation that is both the possibility of performing historiographical research over the ages and the impossibility of focusing the ontology of investigation up front, as the lack of readier-made traces and the ability to decide between competing interpretations will push the historian farther and farther down the causal chain to refute some of the interpretations (and herein lies the underdetermination of historical reconstruction, as Tucker pointed out) and strengthen the others (where the notion of strengthening may be related to Lewis' similarity metric, a point that needs further analysis). Here Lewis' points of world become relevant again.
