tion or thought is analysed into its several elements, and these are set over against one another as so many inde pendent words. The relations of grammar are consequently denoted by position, the particular position of two or more words determining the relation they bear to each other. The analysis of the sentence has not been carried so far in agglutinative languages like Turkish. In these the relations of grammar are represented by individual words, which, however, are subordinated to the words expressing the main ideas intended to be in relation to one another. The de fining words, or indices of grammatical relations, are, in a large number of instances, placed after the words which they define ; in some cases, however, as, for example, in the Ba-ntu languages of southern Africa, the relation is con ceived from the opposite point of view, the defining words being prefixed. The inflexional languages call in the aid of a new principle. The relations of grammar are denoted symbolically either by a change of vowel or by a change of termination, more rarely by a change at the beginning of a word. Each idea, together with the relation which it bears to the other ideas of a proposition, is thus represented by a single word ; that is to say, the ideas which make up the elements of a sentence are not conceived severally and independently, as in Chinese, but as always having a certain connexion with one another. Inflexional languages, how ever, tend to become analytical by the logical separation of the flexion from the idea to which it is attached, though the primitive point of view is never altogether discarded, and traces of flexion remain even in English and Persian. In fact, there is no example of a language which has wholly forsaken the conception of the sentence and the relation of its elements with which it started, although each class of languages occasionally trespasses on the grammatical usages of the others. In language, as elsewhere in nature, there are no sharp lines of division, no sudden leaps ; species passes insensibly into species, class into class. At the same time the several types of speech polysynthetic, isolating, agglutinative, and inflexional remain clear and fixed; and even where two languages belong to the same general type, as, for instance, an Aryan and a Semitic language in the inflexional group, or a Kaffre and a Turkish language in the agglutinative group, we find no certain example of grammatical interchange. A mixed grammar, in which the grammatical procedure of two distinct families of speech is intermingled, is almost, if not altogether, unknown. It is obvious, therefore, that grammar constitutes the surest and most important basis for a classification of lan guages. Words may be borrowed freely by one dialect from another, or, though originally unrelated, may, by the action of phonetic decay, come to assume the same forms, while the limited number of articulate sounds anci conceptions out of which language was first developed, and the similarity of the circumstances by which the tirst speakers were every where surrounded, naturally produce a resemblance between the roots of many unconnected tongues. Where, however, the fundamental conceptions of grammar, and the machinery by which they are expressed are the same, we may have no hesitation in inferring a common origin.
>rms of The main results of scientific inquiry into the origin and van primitive meaning of the forms of Aryan or Indo-European
- " grammar may be summed up as follows. We start with
stems or themes, by which are meant words of two or more syllables which terminate in a limited number of sounds. These stems can be classed in groups of two kinds, one in which the groups consist of stems of similar meanings and similar initial syllables, and another in which the final syllables alone coincide. In the first case we have what are termed roots, the simplest elements into which words can be decomposed ; in the second case stems proper, which may be described as consisting of suffixes attached tc roots. Roots, therefore, are merely the materials out of which speech can be made, the embodiments of isolated conceptions with which the lexicographer alone has to deal, whereas sterns present us with words already combined in a sentence and embodying the relations of grammar. If we would rightly understand primitive Aryan grammar, we must con ceive it as having been expressed or implied in the suffixes of the stems, and in the order according to which the stems were arranged in a sentence. In other words, the relations of grammar were denoted partly by juxtaposition or syntax, partly by the suffixes of stems. These suffixes were probably at first unmeaning, or rather clothed with vague significations, which changed according to the place occupied in the sentence by the stem to which they were joined. Gradually this vagueness of signification disappeared, and particular suffixes came to be set apart to represent particular relations of grammar. What had hitherto been expressed by mere position now attached itself to the terminations or suffixes of stems, which accord ingly became full-grown words. Some of the suffixes denoted purely grammatical ideas, that is to say, were flexions ; others were classificatory, serving to distinguish nouns from verbs, presents from aorists, objects from agents, and the like ; while others, again, remained unmeaning adjuncts of the root. This origin of the flexions explains the otherwise strange fact that the same suffix may sym bolize wholly different grammatical relations. In Latin, for instance, the context and dictionary will alone tell us that mus-as is the accusative plural of a noun, and am-as the second person singular of a verb, or that mus-a is the nomin ative singular of a feminine substantive, bon-a the accusa tive plural of a neuter adjective. In short, the flexions were originally merely the terminations of stems which were adapted to express the various relations of words to each other in a sentence, as these gradually presented themselves j to the consciousness and were extracted from what had been j previously implied by position. Necessarily, the same suffix might be used sometimes in a classificatory, some times in a flexional sense, and sometimes without any definite sense at all. In the Greek dative-locative Tro S-eo-o-i, for example, the suffix -es is classificatory ; in the nomina tive 7rd8-e9, it is flexional. When a particular termination or suffix once acquired a special sense, it would be separated in thought from the stem to which it belonged, and attached in the same sense to other stems and other terminations. Thus in modern English we can attach the suffix -ize to almost any word whatsoever, in order to give the latter a transitive meaning, and the Greek TrdSeo-cri, quoted above, really contains no less than three suffixes, es, -<rv, and -i, the last two both denoting the locative, and coalescing, through cr/t, into a single syllable -en. The latter instance shows us how two or more suffixes denoting exactly the same idea may be tacked on one to another, if the original force and significa tion of the first of them comes to be forgotten. Thus, in Old English sany-estre was the feminine of sang-ere, "singer," but the meaning of the termination has so entirely died out of the memory that we have to add the Romanic -ess to it if we would still distinguish it from the masculine singer. A familiar example of the way in which the full sense of the exponent of a grammatical idea fades from the mind and has to be supplied by a new exponent is afforded by the use of expletives in conversational English to denote the superlative. " Very warm " expresses little more than the positive, and to represent the intensity of his feelings the Englishman has recourse to such expressions as "horri bly warm" like the German "schrecklich warm." Such words as "very," "horribly," "schrecklich," illus trate a second mode in which Aryan grammar has found
means of expression. Words may lose their true significa-
