lly abundant for the different divisions of the
subject. The result is open to doubt, unless we are sure that the
portion to which enumeration was applied was exactly similar to the
remainder.

(4) _Sampling_ is a process of enumeration restricted to a few units
taken at different points in the field of investigation; we calculate
the proportion of cases (say 90 per cent.) where a given characteristic
occurs, we assume that the same proportion holds throughout, and if
there are several categories we obtain the proportion between them. In
history this procedure is applicable to facts of every kind, for the
purpose of determining either the proportion between the different forms
or usages which occur within a given region or period, or the proportion
which obtains, within a heterogeneous group, between members belonging
to different classes. This procedure gives us an approximate idea of the
frequency of facts and the proportion between the different elements of
a society; it can even show what species of facts are most commonly
found together, and are therefore probably connected. But in order that
the method may be employed correctly it is necessary that the samples
should be representative of the whole, and not of a part which might
possibly be exceptional in character. They should therefore be chosen at
very different points, and under very different conditions, in order
that the exceptions may compensate each other. It is not enough to take
them at points which are _distant_ from each other; for example, on the
different frontiers of a country, for the very circumstance of
situation on a frontier is an exceptional condition. Verification may be
had by following the methods by which anthropologists obtain averages.

(5) _Generalisation_ is only an instinctive process of simplification.
As soon as we perceive a certain characteristic in an object, we extend
this characteristic to all other objects which at all resemble it. In
all human concerns, where the facts are always complex, we make
generalisations unconsciously; we attribute to a whole people the habits
of a few individuals, or those of the first group forming part of the
people which comes within our knowledge; we extend to a whole period
habits which are ascertained to have existed at a given moment. This is
the most active of all the causes of historical error, and one whose
influence is felt in every department, in the study of usages and of
institutions, e