than the rest, and who tell you, "I only believe in the
eloquence of figures." Such people do not realise that battalions of
figures are like battalions of men, not always so strong as is supposed.

However, Professor Hyslop took all the "incidents" or statements made
by the communicators and classed them according to the amount of truth
or error they contained. He then divided the incidents into factors. I
will give an example which will help me to define later on what
Professor Hyslop means by _incident_ and _factor_[82]: "My Aunt Susan
visited my brother." This is an incident, or statement of a complete
fact. This incident is composed of four factors which are not
necessarily connected with one another. The first is _my aunt_, the
second the name _Susan_, the third the _visit_, the fourth _my brother_.
Therefore an incident may be defined as a name, a conception or a
combination of conceptions forming an independent fact; it may be again
a combination of possibly independent facts forming a single whole in
the mind of the communicator. The factors would be the facts, names,
actions, or events which do not necessarily suggest each other, or which
are not necessarily suggested by a given name or fact.

Naturally, in tables constructed on these lines, the facts cannot be
classified according to their importance as _proofs_; they can only be
reckoned as true or false. Thus incidents which have only a restricted
value as proofs are on a level with others which are in themselves very
valuable as proofs. This is really the weak point of these statistics.
The proofs need to be examined one by one, and not as a whole.

However, the tables have one advantage; the greatest sceptic, after a
glance at them, can no longer invoke chance, the great _Deus ex machina_
of the ignorant and indolent.

Professor Hyslop has constructed a table for each sitting, and a table
of the sittings as a whole. I cannot reproduce these tables for the
readers, who would require the notes of the sittings to understand them.
I shall only give the definite results.

Thus, out of 205 incidents, 152 are classed as true, 37 as
indeterminate, and only 16 as false. Out of the 927 factors composing
these incidents, 717 are classed as true, 167 as indeterminate, and 43
as false.[83]

It should be said that Professor Hyslop has perhaps overestimated the
number of false and unverifiable incidents. Many incidents or factors
classed as false or unverifiable h