First, it conduces to brevity and clearness to have separate names for the oft-recurring combinations of feelings; but, as these can be defined without reference back to the feelings themselves, it is _enough_ for a _descriptive_ terminology, if there be a name for every variety of elementary feeling, since none of these can be defined, or indicated to a person, except either by his having the sensation itself, or being referred through a known mark to his remembrance of it. The meaning of the name when given to a feeling is fixed, in the first instance, by convention, and must be associated _immediately_, not through the usage of ordinary language, with the feeling, so that it may at once recall the latter. But even among the elementary feelings, those purely mental, and also sensations, such as those from disease, the identity of which in different persons cannot be determined, cannot be exactly _described_. It is only the impressions on the outward senses, or those inward feelings connected uniformly with outward objects (and, consequently, sciences, such as botany, conversant with outward objects), which are susceptible of an exact descriptive language. Secondly, there must also be a separate name for every important common property recognised through that comparison of observed instances which is preparatory to induction (including names for the classes which we artificially construct in virtue of those properties). For, although a definition would often convey the meaning, both time and space are saved, perspicuity promoted, and the attention excited and concentrated, by giving a brief and compact name to each of the new _general conceptions_, as Dr. Whewell calls them, that is, the new results of abstraction. Thenceforward the name nails down and clenches the unfamiliar combination of ideas, and suggests its own definition. Thirdly, as, besides the artificial classes which are marked out from neighbouring classes by definite properties to be arrived at by abstraction, there are classes, viz. _kinds_, distinguished severally by an unknown multitude of independent properties (and about which classes therefore many assertions will be made), there must be a name for every _kind_. That is, besides a terminology, there must be a nomenclature, i.e. a collection of the names of all the lowest _kinds_, or _infimæ species_. The Linnæan arrangements of plants and animals, and the French of chemistry, are nomenclatures. The peculiarity of a name which belongs to a nomenclature is, not that its meaning resides in its denotation instead of its connotation (for it resides in its connotation, like that of other concrete general names); but that, besides connoting certain attributes which its definition explains, it also connotes that these attributes are distinctive of a _kind_; and this fact its definition cannot explain.