ated as the gas that constitutes 4/5 of the atmosphere.

Sec. 4. The rules for testing a Definition are: I.--As to its Contents--

(1) It must state the whole connotation of the name to be defined.

(2) It must not include any quality derivative from the connotation.
Such a quality is called a Proprium. A breach of this rule can do no
positive harm, but it is a departure from scientific economy. There is
no need to state in the definition what can be derived from it; and
whatever can be derived by causation, or by mathematical demonstration,
should be exhibited in that manner.

(3) It must not mention any circumstance that is not a part of the
connotation, even though it be universally found in the things denoted.
Such a circumstance, if not derivable from the connotation, is called an
Accident. That, for example, the lion at present only inhabits the Old
World, is an accident: if a species otherwise like a lion were found in
Brazil, it would not be refused the name of lion on the score of
locality. Whilst, however, the rules of Logic have forbidden the
inclusion of proprium or accident in a definition, in fact the
definitions of Natural History often mention such attributes when
characteristic. Indeed, definitions of superordinate classes--Families
and Orders--not infrequently give qualities as generally found in the
subordinate classes, and at the same time mention exceptional cases in
which they do not occur.

II.--As to its Expression--

(4) A Definition must not include the very term to be defined, nor any
cognate. In defining 'lion' we must not repeat 'lion,' nor use
'leonine'; it would elucidate nothing.

(5) It must not be put in vague language.

(6) It must not be in a negative form, if a positive form be obtainable.
We must not be content to say that a lion is 'no vegetarian,' or 'no
lover of daylight.' To define a curve as a line 'always changing its
direction' may be better than as 'in no part straight.'

Sec. 5. The process of determining a Definition is inseparable from
classification. We saw that classification consists in distributing
things into groups according to their likenesses and differences,
regarding as a class those individuals which have most qualities in
common. In doing so we must, of course, recognise the common qualities
or points of likeness; and to enumerate these is to define the name of
the class. If we discover the qualities upon which a class is based by
direct observation