is real, at least one of the statements is false.
In such cases it is a natural tendency to seek to reconcile them by a
compromise--to split the difference. This peace-making spirit is the
reverse of scientific. A says two and two make four; B says they make
five. We are not to conclude that two and two make four and a half; we
must examine and see which is right. This examination is the work of
criticism. Of two contradictory statements, it nearly always happens
that one is open to suspicion; this should be rejected if the competing
statement has been judged very probably true. If both are open to
suspicion, we abstain from drawing any conclusion. We do the same if
several statements open to suspicion agree together as against a single
statement which is not suspected.[172]

V. When several statements agree, it is still necessary to resist the
natural tendency to believe that the fact has been demonstrated. The
first impulse is to count each document as one source of information. We
are well aware in matters of every-day life that men are apt to copy
each other, that a single narrative often serves the turn of several
narrators, that several newspapers sometimes happen to publish the same
correspondence, that several reporters sometimes agree to let one of
their number do the work for all. We have, in such a case, several
documents, several statements--have we the same number of observations?
Obviously not. When one statement reproduces another, it does not
constitute a new observation, and even if an observation were to be
reproduced by a hundred different authors, these hundred copies would
amount to no more than one observation. To count them as a hundred would
be the same thing as to count a hundred printed copies of the same book
as a hundred different documents. But the respect paid to "historical
documents" is sometimes stronger than obvious truth. The same statement
occurring in several different documents by different authors has an
illusory appearance of multiplicity; an identical fact related in ten
different documents at once gives the impression of being established by
ten agreeing observations. This impression is to be distrusted. An
agreement is only conclusive when the agreeing statements represent
_observations_ which are independent of each other. Before we draw any
conclusion from an agreement we must examine whether it is an agreement
between _independent_ observations. Two operations are thus req