under some combination of them. Consider but one of these attempts, that which seeks to explain the mutual gravitation of any two bodies by the assumption that space is filled by innumerable solid particles in incessant motion, and that the two bodies by mutually shielding each other, tend, so to say, to be pounded towards each other.

That such an explanation, if it be true, lies within the ideal of science is of course indisputable; and when it is established we shall have made a further step in simplification. But this is not quite the question now before us. The point here is whether the fact that such a result would effect a simplification gives any à priori evidence in favour of the theory: If true, the theory would serve to simplify matters, but does this make it more likely to be true? The opinion which on the whole at present occupies the field in regard to Gravitation is that which regards it as an ultimate fact that all bodies, when left free, tend to approach to each other with assignable acceleration. Against such a supposition, however, many minds violently rebel; but except from a natural wish to simplify, it is not easy to see on what exact ground. The mere objection that such an ultimate law is 'inexplicable' is surely invalid: precisely the same may be said against the three laws of motion; and a slight diminution in the number of our inexplicable assumptions is no very weighty motive.

It may be urged that the grounds for accepting the Darwinian Hypothesis were of much the same character. The doctrine of separate creations for the fauna and flora of each separate Ocean Island was long accepted as an ultimate fact, but this gave way at once before the suggestion of an explanation which resolved the many distinct suppositions into various cases of one group of causal influences. The analogy here however does not seem a very fair one, because the simplification was of such a colossal kind. It did not merely concern the reduction of a small number of agencies to one somewhat smaller, but it made a clean sweep of a simply indefinite number; for the natural demands of those who claimed separate creations for each species, and areas of creation for each locality, had a tendency to grow without limit as our knowledge of the world present and past was extended. And in addition to this the invocation of continual creations was the introduction, not of

one more assumption, but one of a kind which was sui generis. A fairer analogy to the doctrine formerly prevalent about the Origin of Species would be found in supposing that men had once taken it for granted that every shooting star was just created at the moment it first came into view. Such a theory would not find much support, as against one which suggested that known laws of motion and matter would account for the sudden flaring up of meteors when they came into contact with the atmosphere of the earth.

As regards then the presumption in favour of simplification it seems to me that we can say so much as this. Given a certain set of facts it is our obvious duty to simplify the arrangement of them as much as possible,-as we shall see again presently when we come to deal with classification,-but this concerns our way of viewing the facts rather than their actual nature. Where we are able however to effect a very great simplification by some alternative supposition, and especially when we thus get rid of what we call independent acts of creation, there is a strong presumption in favour of so doing. But when the difference amounts to nothing more than the ability to get along with an ultimate law or two the less, the presumption in favour of the truth (as distinguished from the convenience) of such a supposition, seems to me to be comparatively small.

Two examples, one which has long been recognized as scientific, and the other of a class which has only just begun to claim such recognition,-may serve to explain what is here meant. The first is that of Gravitation. One view is that we must admit as an ultimate fact that all bodies tend to move towards each other; that there is nothing between them or behind them to make them do so. When to this view is opposed the counter supposition that the movement is caused by the pelting of atoms from behind, or the communication of pressure or vibration through an intervening fluid, I quite recognize that, given the existence of such media and assuming that the facts would be equally well accounted for in such a way, it is desirable to adopt such a simplification. It is our scientific duty to do so. But I cannot see that the fact that we should thus furnish nature with one set of ultimate laws the less is itself a serious argument in favour of either theory.

The other example is that of the so-called Telepathy, as described and supported in the Phantasms of the Living: the doctrine namely that one mind can operate in certain cases directly upon another without the intervention of any physical medium. (I am not arguing in favour of this hypothesis, which seems to me not yet to have become, strictly speaking, probable.) Here, as before, it does not seem to me that the mere difference between one more or less of ultimate postulates or laws counts for much. We must suppose, of course, that the facts have, as regards their broad outlines, been established beyond doubt, and beyond reasonable explanation from chance coincidence; and the alternative will then be between assuming that it is an 'ultimate law' that one mind will tend to induce a similar state (under certain very special circumstances) in certain other minds, and assuming that the observed coincidence or sequence is a result of various other laws', such as fraud, hallucination, and so forth. The consequent detriment of a mere addition of another ultimate law to those (say) of mental association, or of physical and physiological sequence, does not seem to me to count for much.

Turn now to Classification. After the long chapter already devoted to this subject we need not say much about its ideal character; partly indeed because, as will be inferred from what was there said, we are so little competent at present to forecast its future. We saw that the principal improvement already effected consisted in the substitution of a 'Natural' for an 'Artificial' classification; but we also saw that when we endeavoured to bring to light the real distinction between these systems several accounts might be given. It might be said that what was aimed at was the arrangement of the species in accordance with their actual genealogical affinity. If this account were adopted, the ideal of classification would become a History of Creation',- such as, in fact, Hæckel has already attempted to sketch out. If however we preferred the account adopted by Mill, that the object of classification is to arrange the objects in such a way that each class shall comprise the greatest number of common attributes, regard being had to the importance of these attributes; it seems by no means clear what would be the final outcome of such a process. Importance is a very fugitive thing, depending upon the habits, studies

and tastes of men, and on these points no estimate by the present, for the future, is worth much.

It seems therefore preferable to regard Classification as a means to an end, viz. to the effective study and control of the objects in question, and to anticipate that the final outcome of it will not so much take the form of some one arrangement universally accepted, as of a continually increasing number of arrangements devised to meet the ever extending wants of students. An illustration may be gathered from a comparatively trivial subject,—Genealogy,-purposely chosen for that reason, because the very shallowness and artificiality of the subject-matter enables us to look with some confidence further ahead than if we were dealing with the profound problems underlying the biological and social studies.

Suppose, for instance, we have a large number of names of inhabitants of some county, or students of some school or college. What we want, of course, is complete and ready control of this mass of material; that is, given any characteristic whatever, we wish to determine at the shortest notice which of the objects possess this and which do not. For this purpose we clearly want a very large number of arrangements. We should begin with an alphabetical one, because this is the readiest means of picking out a person by his name. We should want a chronological order, for we might be asked who were the contemporaries of such and such a man, and here the alphabet would not help at all. We should want a topographical order, so that those interested in a given place might know what men were connected with that place. We should want the men arranged in accordance with their professions, so that the historian of the church or the bar could pick out his representatives without the labour of working through the entire list. We should want them grouped by their dates of death and places of burial, by the localities where their wills were proved, by such facts as where and to whom they were married, and so on. In a word, nothing would represent finality here except the attainment of a distinct arrangement in accordance with every distinct characteristic which we had occasion to consider. Our only way of securing completeness would be, not to make a compromise amongst these various attributes so as to secure the best possible single arrangement, but rather to

aim at a complete assortment of separate arrangements each involving only one characteristic.

I cannot but think that this will be the tendency in more serious branches of study. Take the case of Botany. It is surely unlikely that the systematists will go on for ever endeavouring to put the finishing touches to the so-called Natural System. Those who want to study genealogical affinity will prefer to do so directly, without the constraint involved in the traditions of classification. And the innumerable demands of other students will have to be met by separate arrangements, each devised for their sole satisfaction. The topographical botanist will have his arrangement, such as H. C. Watson attempted in his Cybele Britannica,-in which local distribution will be the determining attribute. The pharmacologist will have his arrangement; the commercial man his, and so on. There seems no practical limit to the number of such classifications, whether they go by that name, or are termed Indices or Dictionaries, or what not,-which would be really serviceable in the great objects of all classification, viz. that of saving time in looking for something we want and keeping clear control over a multitude of details. The ideal therefore would be sought in a continual addition to these labour-saving resources; that is, in the accumulation of special classifications rather than in the attempt to agree in the construction of some one dominant and exclusive arrangement.

II. So far we have dealt mainly with qualitative considerations. When we come to those which are quantitative we feel on somewhat clearer ground as regards our general aim and the extent to which we may hope gradually to secure it. What we want, in a word, is to "take out the quantities" in respect of all the phenomena of nature. This seems to lead to two subdivisions.

(i) In the first place we want a complete set of units or standards. Those in present use are not only occasionally illchosen in themselves, but there is not as yet universal agreement in their acceptance. The resources of scientific inference are clearly imperfect until these defects are remedied. Then again there is still an immense deal to do in the way of connecting the various units with each other, so as to express one in terms of another. Two familiar examples will serve to show