Given the sun,the planets,and gravitation,we can trace the whole past and future history of the solar system;but the facts at some period must be 'given.'We cannot say that they must,but only that they do exist.Mill himself puts this(75)with all desirable clearness.He expresses it by saying that besides 'causation'there is 'collocation,'a word,he says,suggested by Chalmers.(76)To know the 'collocation,'therefore,is essential.A 'law'does not tell us that there 'must'be plums and suet,but only that if there are such things in certain 'collocations'a plum pudding 'must'be the result.All statements of fact have thus an empirical basis.This,however,takes a peculiar turn in his exposition,and one which is characteristic of a Utilitarian failing.He makes the distinction of relations correspond to a distinction of things.Instead of saying that both causation and collocation are implied in all phenomena,he speaks of some 'uniformities'as dependent upon causation and others as dependent upon collocation.He therefore writes a chapter on 'uniformities of coexistence not dependent on causation.'(77)This,however,is closely connected with,and must be explained by,another doctrine to which he attached the highest importance.After telling us how he was started afresh by Stewart's account of axioms,he adds that he came to 'inextricable difficulties'in regard to induction.He had come to the 'end of his tether,'and 'could make nothing satisfactory of the subject.'When,after five years'halt,he again set to work,he introduced his 'theory of kinds,'which,as he intimates,got round the difficulty.He felt,as we may conjecture,that he had now reduced all the facts to such purely empirical conjunctions that he did not see how to get any tie between them.Any cause,so far as we have gone,might lead to any effect.and even when we have seen a case of conjunction,we can give no reason for its recurrence.Induction enables us to predicate attributes of a class;but a logical class is itself merely a bundle of attributes arbitrarily selected,and it remains to see why,from a thing's possession of some of the class attributes,we can infer that it has the others.Why should not the same set of attributes form part of different bundles?and if so,what is the justification for the primary logical procedure?From featherless bipedism we infer mortality.But why may not some class of featherless bipeds be immortal?If we admit the possibility,all induction becomes precarious.The 'theory of kinds'was,it seems,intended as an answer to these obvious difficulties.
VI.REAL KINDS
Mill's account of 'real kinds'corresponds,as he tells us,to the old logicians'distinction between genus and species.
Though our classification may be arbitrary and nothing properly deducible from it,except the mere fact that we have chosen to give names to certain clusters of attributes,there is also a real difference.Some of our classes do not correspond to 'real kinds,'and are mistaken for them.Others,however,correspond to a real or natural kind.The difference is this:a 'real kind'has an 'indeterminate multitude of properties,not derivable from each other,'whereas an arbitrary or merely logical 'kind'may only differ in respect of the particular attribute assigned.
Thus,to say that Newton is a man is to attribute to him the 'unknown multitude of properties'connoted by 'man.'To say that Newton is a Christian is only to attribute to him a particular belief and whatever consequences may follow from having that belief.(78)One classification,as he says,'answers to a much more radical distinction in the things themselves,than the other does';and a man may thus fairly say,if he chooses,that one classification is made 'by Nature'and one 'by ourselves,'provided that he means no more than to express the distinction just drawn.Now,it is easy to understand why Mill felt that this assertion entitled him to a 'real'bond which would keep phenomena together in a more satisfactory way.All things had become so loose and disconnected that it was difficult to explain any extension of knowledge even by induction.Yet,whatever the reason,things do stick together in coherent and many-propertied clusters.The bond seems to be real when it is stated 'objectively,'not 'subjectively'--as a property of the things observed,not of the classes made by the mind itself.I take the remark to be both true and important;and,moreover,that Mill deserves credit for perceiving so clearly this weak joint in his armour.His application,however,suggests,when he had hit upon an apparent escape from his 'inextricable'difficulties,he was too much relieved to work out its full effect upon his general theory.The 'theory of kinds'is inserted rather than embodied in his philosophy,and makes rents in the attempt to fill a gap.It plays,however,so important a part in the doctrine that it requires some further consideration.
A real kind,we see,has two characteristics;it has innumerable properties,and those properties are not 'derivable'from the others.In fact,a derivative property would be merely a modification of a primitive property.A geometrical 'kind,'a curve of the second order,for example,has innumerable properties,but they are all derivative from the simple properties expressed in the axioms and definitions.They reciprocally imply each other.But can we say the same of the properties of a thing --of a plant or of water or of an atom?
