False in one particular, false in everything. This principle of Roman law is still respected and has been appropriated by other disciplines. The concept is that if a witness has been shown to lie in one particular respect in a case, he is not to be trusted in anything else he says. This is why it is important for attorneys to impeach opposing witnesses in court: it discredits the rest of their testimony. The object behind the principle is to reject questionable testimony (even if it might be true) before accepting falsehood into evidence.

The legal principles of interrogating witnesses have been drawn into the task of evaluating historical sources. Just as a witness in court can be impeached by being shown to have lied, an historical source likewise loses much of its authority if its author can be shown to have deliberately falsified something--how can we trust an author concerning fact X when we know him to have lied about fact Y? Such an author may corroborate something a better witness says, but has forfeited our trust where he speaks without corroboration.

So, too, a manuscript bearing copies of ancient works is called a witness: not to a crime, obviously, nor to a contract, nor to historical facts, but rather to an earlier version of its text. Many of the same principles have been drawn into this field as well. A manuscript which contains many errors or bad readings (for example, a simpler phrase replacing a more difficult one which the scribe did not understand, or frequent spelling blunders) cannot be trusted without corroboration from an independent manuscript (i.e., one which is neither its copy nor its descendent).

A related concept is testis unus, testis nullus (one witness is no witness), q.v. This is what gives the impeachment of corroborating witnesses its bite. If you have one witness of something, you are mired in the old "he said, she said" problem. This principle is weaker in historical source criticism, because sometimes we have only one source, and we have to make the best of it. In the study of manuscripts, one independent witness often summons forth an editor's attempts to amend troubled patches of text with superior readings. This is at best a plausible guess, though papyrus discoveries occasionally confirm such speculation.