LANGUAGES.] PHILOLOGY 783 been split in two, but this process again has taken place with perfect regularity, the one sound appearing before r, the other before all other consonants. It is easy to see that the common practice of comprising the history of the Old English d in the one rule, that it was changed into the sound of the a in make except when followed by an r, can only be defended on the practical ground that this rule is convenient to remember, because the words exhibiting the former change are more numerous than the instances of the latter ; apart from this there is nothing to justify the assumption that one of these changes is the rule and the other the exception. The fact is, that we have two inde pendent cases of change, which ought to be stated in two distinct and independent rules according to the different positions in which the original d stood before the splitting began. It is also easy to observe that the variety of modifying influences may be much more manifold than in the present instance of make and hare, and that the number of special phonetic rules in such cases must be increased in proportion to the progress made in the investigation of the said modifying powers. In this respect much still remains to be done, but what has been achieved is more than sufficient to prove the correctness of the statement from which we started above, that phonetic rules in them selves are without exceptions, however often phonetic processes may have been crossed and more or less effaced by non-phonetic influences in actual (especially literary) language, such as mixture of dialects, formation by analogy, and the like. Analogical change, on the other hand, does not affect the pronunciation of a language as a whole in the way phonetic change does, but is confined to the formation and inflexion of single words or groups of words, and therefore very apt to bear an entirely arbitrary and irregular character. A few instances w r ill be sufficient to illustrate this. In Old English a certain number of substantives formed their plurals by mutation of the root-vowels, as/of fet or boc bee. In Modern English this system of inflexion has been preserved in some cases, as in foot feet, and altered in others as book books. Now, while foot, feet, and book are the regular modern phonetic equivalents of the old fot, fet, boc, the plural books can in no way be phonetically traced back to the old bee, the phonetical equivalent of which in Modern English would be *beech. The only possible ex planation of a form like books is that the older bee was at some date given up and replaced by an entirely new forma tion, shaped after the analogy of the numerous words with a plural in -s without modification of the root-vowel. That this should have been done in the case of book, but not in that of foot, is an accident, which must be accepted as a fact not allowing of any special explanation. Let us now take another instance from the English verb. In Old English the different persons of the preterite indicative in the so-called strong (irregular) verbs were generally dis tinguished by different root-vowels ; ridan, " to ride," and bindan, "to bind," for instance, form their preterites thus : ie rdd, Gw ride, he rdd, we, ge, hie ridon, and ic band, <Sti bunde, he band, we, ge, hie bundon. In Modern English this difference in the root-vowels has been abandoned, and rode, bound now stand for all persons, rode being the modern phonetic equivalent of the 1st and 3d sing, rdd, while bound represents the w-forms of bindan. Inasmuch as a similar process of levelling has been carried through in all preterites of Modern English, regularity prevails even here. But when we look to its results in the indi vidual verbs we soon find that the choice amongst the different forms which might have served as starting-points has been entirely arbitrary. It is indeed impossible to say why the old singular form should have been chosen as a model in one case, as in rode, and the old plural form in another, as in bound. From these and numerous similar instances we must draw the conclusion that it is beyond our power to ascertain whence analogical changes may start, and to what extent they may be carried through when once begun. All we can do is to carefully classify the single cases that come under our observation, and in this way to investigate where such changes are especially apt to take place and what is their general direction. As to the latter points, it has been observed before that levelling of existing differences is one of the chief features in ana logical change (as in the case of rode and bound}. As to the former, it must be borne in mind that, before any analogical change can take place, some mental connexion must exist between the words or forms serving as models and those which are remodelled after the types suggested to the mind of the speakers through the former. Of such natural mental combinations two classes deserve especial notice : the mutual relationship in which the different, say in flexional, forms of the same word stand to each other, and the more abstract analogies between the inflexional systems of word-groups bearing a similar character, as, for instance, the different declensions of nouns and pronouns, or the different conjugations of verbs. The instance of rode, bound may serve to illustrate the former category, that of books the latter. In the first case a levelling has taken place between the different forms of the root-vowels once exhibited in the different preterite forms of ridan or bindan, which clearly constitute a natural group or mental unity in consequence of their meaning. The form of rode as a plural has simply been taken from the old singular, that of bound as a singular from the old plural. In the case of book books for boc bee, this explanation would fall short. Although we might say that the vowel of the singular here was carried into the plural, yet this would not explain the plural -s. So it becomes evident that the old declension of boc bee was remodelled after the declen sion of words like arm arms, which had always formed their plurals in -s. Isolated words or forms, on the other hand, which are no part of natural groups or systems, inflexional or formative, must be regarded as commonly safe from alterations through analogy, and are therefore of especial value with regard to establishing rules of purely phonetic development. It is true that the distinction between phonetic and analogical change has always been acknowledged in com parative philology. At the same time it cannot be denied that analogical changes were for a long time treated with a certain disdain and contempt, as deviations from the only course of development then allowed to be truly " organic " and natural, namely, that of gradual phonetic change (hence the epithet " false " so constantly attached to analogy in former times). Amongst those who have recently contributed most towards a more correct evalua tion of analogy as a motive -power in language, Professor Whitney must be mentioned in the first place. In Ger many Professor Scherer (Zur Geschichte der deutsehen Spraehe, 1868) was the first to apply analogy as a prin ciple of explanation on a larger scale, but in a wilful and unsystematic way. Hence he failed to produce an immediate and lasting impression, and the merit of having introduced into the practice of modern comparative philo logy a strictly systematic consideration of both phonetic and analogic change as co-ordinate factors in the develop ment of language rests with Professor Leskien of Leipsic, and a number of younger scholars who had more or less The Ne<, experienced his personal influence. Amongst these Brug- School, mann, Osthoff, and Paul rank foremost as the most vigorous and successful defenders of the new method, the correctness of which has since been practically acknow ledged by most of the leading philologists of all shades,
