Chapter 47

CLASSIFICATION, AS SUBSIDIARY TO INDUCTION.

Every name which connotes an attribute thereby divides, but only incidentally, all things, known and unknown, real and imagined, into two classes, viz. those which have, and those which have not the attribute. But sometimes the naming itself is but the secondary and subsidiary, and the classification, the primary object. The general problem of such classification is, to provide that things shall be thought of in such groups, and the groups in such an order, as will best promote the remembrance and ascertainment of their laws. Its subjects are _real things_ exclusively, but _all_ real things, since, to place one object in a group, we ought to know the divisions of nature at large.

Any property may be the basis for a classification; but those best suited are properties which are causes, or, next, as the cause of a class's chief peculiarities seldom serves as its diagnostic, any effect which is a sure mark both of the cause and of the other effects. Only a classification so grounded is scientific; the same also is not technical or artificial, but natural, and emphatically _natural_ (as compared with classifications in an inferior degree also _natural_, which are based on properties important with reference to the reasoner's special practical objects), when the classification is based on those properties which would most impress one who knew all the properties, but was not interested particularly in any one. Further, it is a great recommendation of a classification, that it groups together things of like general aspect; but this is not a _sine quâ non_: a group may be _natural_ even if based on very _unobvious_ properties, provided these are marks of many other properties, though certainly then there should be also some more obvious property to act as a mark of the unobvious ones which form the real basis.

As the first principle of _natural_ classification is that the classes must be so formed that the objects composing each may have as many properties in common as possible to serve as predicates, all _kinds_ should have places among the _natural_ groups, since the common properties of _kinds_, and, therefore, the general assertions that can be made about them, are innumerable. But _kinds_ are too few to make up the whole of a classification: other classes also may be eminently _natural_, though marked out from each other only by a definite number of properties. Of neither sort of _natural_ groups is Dr. Whewell's theory _strictly_ true, viz. that every _natural_ group is not determined by definition, that is, by definite characters which can be expressed in words, but is fixed by Type. He explains that a type is an example of any class, for instance, a species of a genus, which possesses all the characters and properties of the genus in a marked way; that round this type-species are grouped all the other species, which, though deviating from it in various directions and degrees, yet are of closer affinity to it than to the centre of any other group; and that this is the reason why propositions about _natural_ groups so often state matters as being true not in all cases, but only in most. Now, there is a truth, but only a partial truth, in this doctrine. It is this: in forming _natural_ groups, species which want certain of the class-characters, some one, and others another, are classed with those (the majority) that have them all, because they are more like (that is, in fact, have more of the common characters of) that particular group than of any other. On account of the feeling of vagueness hence engendered, we certainly, in deciding if an object belong to the group, do generally (and _must_, when the classification is made expressly with a view to a special inductive enquiry) refer mentally, not as a substitute for, but in illustration of the definition of the group, to some standard specimen which has _all_ the characters well developed. But not the less, therefore, are all _natural_, equally with all artificial, groups framed with distinct reference to certain definite characters. In the case of _kinds_, a few characters are chosen as marks of the rest. In the case of other _natural_ groups, the formation of the larger groups, into which we collect the _infimæ species_, is suggested indeed by resemblance to types (since we form each such larger group round a selected _kind_ which serves as its exemplar); but the group itself, when formed, is determined by definite characters.

Class names should by the mode of their construction help those who have learnt about the thing, to remember it, and those who have not learnt, now to learn, by being merely told the name. This is best effected, in the case of _kinds_, when the word indicates by its very formation the properties it connotes. But this is seldom possible. For, though a _kind_-name connotes not all the _kind_-properties, but some only which serve as sure marks of the rest, even these have been found too many to be included conveniently in a name (except in Elementary Chemistry, where every compound substance has one distinctive index-property, viz. the chemical composition). A subsidiary resource is to point out the _kind's_ nearest natural affinities. For instance, in the binary Nomenclature of Botany and of Zoology, the name of every species consists of the name of the _natural_ group next above, with a word added expressive of some quality in the nature or mode of discovery, or what not, of the particular species itself. By this device (obtaining at present only in Botany and Zoology), as well is the expression, in the name, of many of the _kind's characters_ secured, as the use of names economised, and the memory relieved. Except for some such plan, what hope of naming the 60,000 known species of Plants?