they ought to have gone, and too often generalized tendencies into universal rules. Subsequent writers followed these rules and not the living speech, even though they were writing in what was meant to be a vernacular. Moreover, at an early date, the Prakrits, qua literary languages, began to lose their characteristics as local forms of speech. A writer composed in Maharastri, not because it was his native language, but because it was the particular Prakrit employed for lyrics and in-formal epics. In the same way, in dramatic literature, Sauraséni and Magadhi were put into the mouths of characters in particular walks in life, whatever the nationality of the dramatist might have been. There was thus a tendency for these literary Prakrits to adopt forms from the vernacular dialects of those who wrote them, and, en refuanche, for the very popular lyric poetry of Maharastri to influence the local dialects of the most distant parts of India. On the other hand, although to a certain extent artificial, the literary Prakrits are all based on local vernaculars, a fact entirely borne out by a comparison with the modern Indian languages, which closely agree with them in their mutual points of difference. We now proceed to consider the general points in which the Prakrits differ from Sanskrit and from each other. The reader is throughout assumed to be familiar with the general outline of the article SANSKRIT.
Contractions: Skr.=Sanskrit. Pr.=Prakrit. S.=Sauraseni. Mg. =Magadhi. AMg. =Ardhamagadhi. M. =-Maharastri. Ap. = Nagara Apabhramsa]
Vocabulary.-The vocabulary of S. is to all intents and purposes the same as that of Skr. In the languages of the Outer Band there are numerous provincial words (déii or dé§ ya), the originals of which belonged to Primary Prakrits other than those of the Midland. In the Outer Band there is also a rich variety of grammatical forms, many of which are found in the Veda and not in classical Sanskrit, and some (e.g. Pr. -hi, Pali -dhi, Greek -01.) which cannot be traced to any known Primary Prakrit form, but which must have existed in that stage and beyond it, back into Indo-European times. Phonetics.-The Skr. diphthongs é and 6 are treated in Pr. as pure vowels, and may each be either long or short. Ai and du become either é and 5 or ai and 1112 respectively. The vowel I becomes a, 11, or, under the influence of a neighbouring labial, u. Before two consonants an original long vowel becomes short, and i and u are (according to the grammarians) changed to eand 0 respectively. The last rule is an instance of grammarians' over-generalization, and is not universally true. Examples, Skr. mdrga-, Pr. magga-; Skr. sindura-, Pr. sendum-; Skr. puslaka-, Pr. potthaa-. Conversely, a short vowel before two consonants is lengthened on one of them being elided. Thus, Skr. isvara-, Pr. issara- or Exam-; Skr. jihad, Pr. jihzi. In Ap. the quantity of vowels is very loosely observed. In all dialects n becomes 11 unless it is followed by a dental mute, but in ]aina works nn and initial n remain unchanged. judgirlg from modern vernaculars, the latter seems to have been the real state of affairs. In Mg. j becomes y and r becomes l. Here also r and S become § , a peculiarity still preserved by the modern Bengali. Elsewhere 5 and s usually become 5, but the change of a sibilant to h is not uncommon in the Outer Prakrits (even in Mg.), though rare in the more archaic S.
Initial y becomes j except in Mg., in which, on the contrary, j becomes y. Subject to the foregoing general rules, all other initial consonants usually remain unchanged. As regards medial single consonants:-
I. K, g, c, j, t, d and y are usually elided. As a hiatus is caused by the elision, a faintly sounded y (or in some cases 11) is substituted for the elided consonant, though only written in jaina MSS. Examples: Skr. léka-, Pr. l5(y)a-; Pr. maa=Skr. mata-, mada-, mayamgga or migta-. The latter example illustrates the extraordinary confusion which results from the strict application of this rule of elision of medial consonants. Such a Prakrit would have failed in the main object of a language-the connotation of distinct ideas by distinct sounds. To the present writer it seems impossible that such a language could ever have existed, and he is persuaded that the rule just given is 'merely another instance of grammarians over-generalization. A rule has been made out of a tendency, and this tendency was evidently, first, to soften a hard letter, and then (but not necessarily) to elide it. We see this well illustrated by Apabhrafhfa, in which k, t andg are usually preserved under the forms g, d and b. In the Outer rakrits also k often becomes g, as in Skr. § r61/aka-, Iaina M. and AMg. srivaga-, Mg. .ftivaga-. S. and Mg. always preserve a medial t, changing it to d; thus, Skr. gata-, S. Mg. gada-, elsewhere ga(y)a-.
2. Kh, gh, th, dh, ph and bh similarly become h. Also, as above, S. and Mg. change th to dh. Th becomes dh, and ph may become bh. The other aspirates (ch, jh, dh), and also sometimes bh, remain unchanged. In Ap., as before, kh, th and ph are usually preserved in gh, dh and bh respectively.
3. I becomes d, d becomes Z (often written Z), which when doubled becomes dentalized to ll, as in the case of the jaina nn. P and b usually become v. The Outer languages often cerebralize dental sounds and change ! to Z.
4. N, m, I and h remain unchanged. V disappears before u, but otherwise generally remains unchanged. In Ap. m may become a v nasalized by anumisika; thus, Skr. bhramara-, Ap. bhailara-.
Final consonants usually disappear altogether, except nasals, which become anusmifa. Thus, Skr. samantcit, phalam, Pr. samantti, phalam.
The following rules will be found to include the great majority of possible cases of compound consonants. They show clearly the character of alll changes from Primary to Secondary Prakrit, viz. the substitution, mainly by a process of assimilation, of a slurred for a distinct pronunciation:-
I. In Pr. a conjunct consonant cannot consist of more than two elements, and, except in Mg. and Ap., can only be a double consonant or a consonant preceded by a nasal, a consonant followed by r, or one of the following: nh, nh, mh, Zh. The consonants r and h cannot be doubled.
2. In Pr. the only conjuncts which can begin a word are nh, nh, mh, and lh. If any other conjunct consonant be initial, the first member of the Pr. form of it is dropped. Thus, in Pr. kr becomes kk, and the Skr. dkramati becomes Pr. akkamai. If we omit the initial preposition d- (Pr. a-), the kk becomes initial, and we have liaznai, not *kkamai Similarly, Skr. sthira- becomes Pr. thira- for tt mr.-.
3. 'L and -U are elided when they stand first or last in a compound, and the remaining letter is doubled, if it admits of doubling. Thus, Skr. ulkd, Pr. ukkfi; Skr. pakva, Pr. pakka-. The same rule is followed regarding r, but when it follows a consonant it is sometimes, especially in Ap., retained even when initial. Thus, Skr. arka-, Pr. akka; Skr. priya-, Pr. pia- or (Ap.) pria-4. M, n and y are elided when standing last in a compound, and the remaining letter is doubled; thus, Skr. rafmi-, Pr. rassi-. 5. K, g, lf, d, t, d, p, § , .5 and s are elided when standing first in a compound, and the remaining letter is doubled as before; thus, Skr. bhakta-, Pr. bhatta-; Skr. skhalita-, Pr. khalfia- for *kkhal'ia(see rule 2).
6. The above rules hold in the order given above; that is to say, rule if holds in preference to rules 4 and 5, and rule 4 in preference to ru e 5. Thus, in the Skr. compound kr, the r is elided under rule 3, and not the k under rule 5, so that the Pr. form is kk. 7. Special Rules for Mg.—ln this form of Pr. there are several peculiar changes. Dy, rj, ry, all become yy; ny, ny, jd, ij become hh; medial cch becomes ic; 115, sf, 5th become sgf; and nh, sth become st. Other changes also occur, besides dialectic variations of those given above.
Declensdon.-Pr. has preserved the three genders of Skr., but has lost the dual number. As a rule, the gender of a noun follows that of the Skr. original, though in AMg. there is already a tendency to substitute the masculine for the neuter, and in Ap. these two genders are frequently confused, if the distinction is not altogether neglected. In the formation of cases, the phonetic rules just given are full applied, but there are also other deviations from the Skr. original The consonantal stems which form an important part of Skr. declension are frequently given vocalic endings, and there is a general tendency to assimilate their declension to that of a-bases, corresponding to the first and second declensions in Latin. This tendency is strongly helped by the free use of pleonastic suffixes ending in a, which are added to the base without affecting its meaning. Of these the most common are -ka-, -da-, and -alla-, -1llla- or -ulla-. The first of these was also very common in Skr., but its use became much extended in Pr. In accordance with the general rule, the k is liable to elision; thus, Skr. ghéla-ka-, Pr. ghzida-a-. It may even be doubled, as in Skr. bahu-, much, Pr. bahu-zz-cz-, for bahu-ka-ka-. -Da- is confined to Ap., and may be used alone or together with the other two, as in Skr bdhubala-, strength of arm, Ap. bdhubal-ullada-(k)a-. Illa- is most common in the Outer languages, and especially so in AMg. and M.; thus, Skndpura-, M. pur-illa.-. All the Skr. cases are preserve except the dative, which has altogether disappeared in the Midland, but has survived in the singular number in the Outer languages. Everywhere the genitive can be employed in its place. Most of the case-forms are derived from Sanskrit according to the phonetic rules, but Ap. has a number of dialectic forms which cannot be referred to that language (cf. the remarks above about -hi=6¢). It also rarely distinguishes between the nominative and the accusative. As an example, we may give the commoner forms of the declension of the Skr. putm, Pr. putla-, a son (see next page). It should be understood that numerous other forms were also in use, but the ones given here are selected because they are both common and typical.
The declension of neuter a-bases closely resembles the above, differing only in the nominative and accusative singular and plural. Ap. has almost lost the neuter termination in the singular. Feminine a-stems are declined on the same lines, but the cases have run more into each other, the instrumental, genitive and locative singular