ars after the facts, often at
the very end of the author's career, have introduced innumerable errors
into history. It must be made a rule to treat _memoirs_ with special
distrust as second-hand documents, in spite of their appearance of being
contemporary testimony.

(3) The author states facts which he could have observed, but to which
he did not take the trouble to attend. From idleness or negligence he
reported details which he has merely inferred, or even imagined at
random, and which turn out to be false. This is a common source of
error, though it does not readily occur to one, and is to be suspected
wherever the author was obliged to procure information in which he took
little interest, in order to fill up a blank form. Of this kind are
answers to questions put by an authority (it is enough to observe how
most official inquiries are conducted in our own day), and detailed
accounts of ceremonies or public functions. There is too strong a
temptation to write the account from the programme, or in agreement with
the usual order of the proceedings. How many accounts of meetings of all
kinds have been published by reporters who were not present at them!
Similar efforts of imagination are suspected--sometimes, it is thought,
clearly recognised--in the writings of mediaeval chroniclers.[154] The
rule, then, will be to distrust all narratives conforming too closely to
a set formula.

(4) The fact stated is of such a nature that it could not have been
learnt by observation alone. It may be a hidden fact--a private secret,
for example. It may be a fact relating to a collectivity, and applying
to an extensive area or a long period of time; for example, the common
act of a whole army, a custom common to a whole people or a whole age, a
statistical total obtained by the addition of numerous items. It may be
a comprehensive judgment on the character of a man, a group, a custom,
an event. Here we have to do with propositions derived from observations
by synthesis or inference: the author can only have arrived at them
indirectly; he began with data furnished by observation, and elaborated
them by the logical processes of abstraction, generalisation, reasoning,
calculation. Two questions arise. Does it appear that the author had
sufficient data to work upon? Was he accurate, or the reverse, in his
use of the data he had?

On the probable inaccuracies of an author, general indications may be
obtained from an examination of his