tion of the personal terminations, this modification of the stem served to distinguish mood. Thus when the vowel a was added to a root from which a verb was already formed, the inflexions of this extended root or stem denoted an action intended to be performed, and thus acquired the force of a conjunctive mood. The addition of the vowel having thus obtained this differentiating power, it was afterwards affixed with the same force to stems already provided with a, the contraction of a + a giving d ; hence, just as han-ti is "he kills," han-a-ti, "let him kill," so we get bhara-ti, "he is carrying," bhard-ti, "let him be carrying."
In a fifth stage compound verbal forms make their appearance, i.e., tense-stems are produced by the union of primary verb-stems with the roots of verbs which have become simply auxiliary. That this must have been at a later stage than the preceding processes is clear from the fact that verbs only gradually lose their full meaning, and sink into auxiliaries. The verbs so used are (1) as, originally "breathe," afterwards "be ; (2) ja, "go"; (3) dha, "do." From the composition of the first with a verb-stem we get forms like those of the compound or so-called first aorist e.g., a-dik-sat, "he pointed" ( = e-Set/c-o-e-r). Here we have the union of a root (in this case acting as a noun-stem) with ths auxiliary verbs in the third parson, preceded by the augment ; a-dik-sa-t is to the earlier form a-da-t much as "turn dicens erat" is to "turn dans." These formations belong to the earliest stratum of this period, inasmuch as the stem appears in its simplest form. For a like reason we must assign to the same stratum the compounds of ja with the simple stem. This auxiliary is used to denote relations which were at first somewhat indefinite but were afterwards more precisely differentiated. There is (1) the present of duration: e.g., svid-jd-mi, "I am sweating;" (2) the passive force thus derived, as in Sanskrit ja is a sign of the passive: e.g. bodh-a-ti, " he knows," bodh-jd-te, "he is known;" (3) the tendency to do a thing, i.e., the optative mood : as in as-jd-m, the primitive form of siem (sini) and firjv. There is also to be noted, as belonging to this stage, the very important present of duration from the root as, i.e., as-jd-mi, which acquired for itself, and when affixed to roots or stems gave to them, the force of a future. With regard to the root dha, the widely varying force which its compounds have in the different cognate languages prevents us from determining with certainty the manner in which it was originally employed in composition (cf. Curtius, Das Verbum, ii. 352). To a second stratum of the same period must be assigned those compound verbal formations in which the stem is not a pure root, but has already been developed into a stem which has the character of a noun. If we com pare, e.g., bhdra-jdmi ( = c^ope-Jw-//.^, " I am bearing," with svid-jd-mi, we find in the first a nominal theme employed for composition and inflexion, in the second a simple root. It is of much importance to notice that here too the verbal formation must have preceded the formation of cases. Had the accusative bJtdra-m been in use, it would have been im possible not to employ it in connexion with a verb-form like jd-mi, just as the Romans said venum dare, datum iri, and the like, and as Sanskrit forms the periphrastic perfect of the tenth conjugation, by uniting the auxiliary verb with the accusative e.g., k orajdm k dkdra, bodhajdm babhilva, &c. We are led to the same conclusion by considering forms like a-dik-sa-nt, by which the absence of plural inflexion is not less clearly indicated than the lack of case-inflexions by bhdm-jd-mi. It has been urged, e.g., by Professor Max Miiller, that this argument is a weak one, because our ancestors must have felt the need for clearly distinguishing the plural from the singular, and the nominative from the accusative, before the need for denoting the differences between the persons. To this it may be replied (1) that the argument from what must have been is one of the most dangerous that can possibly be used in philology; conclu sions a priori have again and again been disproved by a more complete acquaintance with the facts of the history of language ; (2) that, as a fact, incompletely developed languages do find it more easy to do without distinct marks of the cases than to dispense with personal inflexions, and that this is confirmed by languages like English and French, which have returned to an uninflected state more com pletely in the case of nouns than in that of verbs ; (3) that in the inflexion of nouns the sign of the plural is added to the case-suffix and not vice versa (e.g., Xwovs = XDKOV-S = varka-m-s), so that the use of the sign of case must have preceded that of the sign of number, although the latter might have seemed to us the more indispensable. Professor Mailer s argument that composition might have taken place in times subsequent to nominal inflexion, because the stem-forms show themselves in certain cases of declension, and therefore might have remained present in the conscious ness of those using the language, breaks down upon the essential distinction between the nature of the composition of the verbal forms in primitive times and of the construc tion of compound verbs within the historical period. We may therefore safely follow Curtius in holding it as at least highly probable that verbs were already inflected according to person, tense, mood, and voice at a time when nouns were still in the state either of simple roots (e.g., rA - ) or nominal themes (bhara, akva, sumi,pati, and the like). The needs of language at this stage were probably helped out (as at present in uninflected languages) by the position of the words, by the stress of the voice, and by a free use of pro nominal roots, which may have been already acquiring somewhat of a prepositional force. But one of the most important means of expression was undoubtedly composi- tion. The form which the elements of compounds take, rarely (and apparently never in any early word) appearing with any case-inflexion of the first element (as in ov Sei/oo-wpa, Surpeff)^, d/x^opecif^opos), but presenting themselves simply as stems (Xoyo-ypa<o-s, &c.), show s that at any rate the mould in which they were cast, the analogy on which later compounds were freely fashioned, was constructed at a time whon nouns were not inflected. The various relations which the factors of the compounds bear to each other point to the same fact. We find in Greek, according to the very clear and careful statement of Curtius (Grammar,^ 359 ; cf. Elucidations, pp. 172-178) three kinds of compounds : (1) determinative, in which the second factor is the principal, which, without altering its meaning, has it defined by the first; e.g., d/cpoTroXi? = aKpa 770X19; (2) attributive, where the first factor defines the second, but so as to alter its mean ing, the two combining to form anew idea ; e.g., yua/cpo-^tp = //.a/cpa<; xcipas t^u>v ; (3) objective, in which one of the two words is grammatically governed by the other, so that in paraphrasing one of the two must be put in an oblique case ; e.y., lyvi-OYO-s = TO, i^via e^wv, <iXo-/xot cro-s = o rag Mot cras <iXu>v, d to-Xoyo-s = Xoyov aios, 6fo-fiafBr]<; = VTTO Oeov /3e/3Xa/x/aeVo9, ^eipo-Troi ^TO-s = X C P" Tonqros. (Simi larly Max Miiller, Sanskrit Grammar, 513, gives six classes of Tatpurusha compounds, according as the first element stands in an accusative, instrumental, dative, ablative, genitive, or locative relation to the second, with numerous examples.) Curtius justly calls attention to the epithet 7rarpo<ov??a as applied in the Odyssey to yEgistheus in re lation to Orestes, " one who had slain his father." It is evident that composition, used as freely as forms like these indicate, could have taken very largely the place of case- inflexions.
The origin of the cases, which marks a sixth stage, presents much more difficulty than the origin of verbal flexion. But one broad division may at once be made. The voca-
