When is a historical source not a historical source?

Just before Christmas I was lucky enough to receive a commission that involved working in the library of the Royal Society and at the Parliamentary Archives. I was new to both of these repositories (both 'real' archives and undeniably memorable places: a cloistered feel, oak-panelled book cases, heavy volumes and all the rest of it), but infinitely patient and helpful staff ensured that my visits were productive. I'm fairly confident that I turned up some interesting material for my client, but the nature of the job set me thinking about a particular - and in some respects slightly troubling - aspect of some of the documentary materials which some historians must use to pursue their craft.

The example I have in mind is the extensive but scattered correspondence of an important eighteenth-century political figure; nothing problematical about that, and in fact surely it is a good thing that so much has survived. Perhaps, but many of these archival remains consist of numerous copies of the same items. I was surprised to find this, although had I bothered to think about it the situation would have been self-evident: an army of clerks turning out copies of copies of copies of an original letter was an obvious and vital component of the machinery of government; cabinet members, lords, dukes, MPs and what not naturally needed copies of important letters and papers to aid their deliberations, but 200 years later how to identify the original and in some circumstances can it even be said to exist? Should one privilege one particular copy or will any of them do? And what if mistakes were introduced at some point in the laborious copying process? Would that compromise the integrity of the historical source? It could be argued that as long as the historian accurately references what he finds in the archives none of this matters as he can only work with what is available, but start thinking too much about this stuff and the discipline can start to look a little shaky. (And this might be relevant in future; think of the vast numbers of copy documents generated by business, government and organisations - which of them can be cited with confidence as the original?)