The lack of a statistical record compels us to base many of our conclusions in matters of ancient history on inference from casual notices. These notices, often obscure, are liable to misinterpretation. Even when rightly interpreted, their authority and their bearing on each other are often doubtful. In particular, there is an ever-present difficulty arising from the loss of other evidence, probably larger in bulk and not inferior in value. The discovery of new evidence sometimes upsets conclusions that at an earlier stage of inquiry were reasonable enough. Re-examination of old evidence often modifies them. But an early conclusion sometimes for lack of sufficient challenge glides into an assumption: if it falls in with the general views of historians, it imperceptibly affects their attitude as inquirers and gains authority by lapse of time.