The problem thus presented to us has some resemblance to that of the restoration of texts of which we have already spoken. In both cases we proceed on the assumption that identical readings have a common source: a number of different scribes, in transcribing a text, will not make exactly the same mistakes in exactly the same places; a number of different writers, relating the same facts, will not have viewed them from exactly the same standpoint, nor will they say the same things in exactly the same language. The great complexity of historical events makes it extremely improbable that two independent observers should narrate them in the same manner. We endeavour to group the documents into families in the same way as we make families of manuscripts. Similarly, we are enabled in the result to draw up genealogical tables. The examiners who correct the compositions of candidates for the bachelor's degree sometimes notice that the papers of two candidates who sat next each other bear a family likeness. If they have a mind to find out which is derived from the other, they have no difficulty in doing so, in spite of the petty artifices (slight modifications, expansions, abstracts, additions, suppressions, transpositions) which the plagiarist multiplies in order to throw suspicion off the scent The two guilty ones are sufficiently betrayed by their common errors; the more culpable of the two is detected by the slips he will have made, and especially by the errors in his own papers which are due to peculiarities in those of his accommodating friend. Similarly when two ancient documents are in question: when the author of one has copied directly from the other, the filiation is generally easy to establish; the plagiarist, whether he abridges or expands, nearly always betrays himself sooner or later.[86]