The other characters speak Prākrit, and, by a curious variation from the normal practice, the stage directions, which are freely given as in the classical drama, are normally expressed in the language which the character concerned uses, though there are cases of mixture and apparent confusion which may be due to the scribe. Three different forms of Prākrit may be distinguished, the first spoken by the Duṣṭa, the second by the mysterious Gobaṁ°, and the third by the hetaera and Vidūṣaka. The Duṣṭa’s speech in three important points is similar to the Māgadhī of the Prākrit grammarians; it substitutes l for r; reduces all three sibilants to ç; and has e in the nominative singular of masculine nouns in a. But it ignores the rules of the grammarians in certain matters; hard letters are not softened (e.g. bhoti), nor soft consonants elided (e.g. komudagandha), when intervocalic. There is no tendency to cerebralize n, and in kālanā the dental replaces the cerebral. Fuller forms of consonants remain in han̄gho (haṅho) and bambhaṇa (bamhaṇa). The later forms of development of consonantal combinations are unknown; thus for rj we have jj, not yy, as in ajja; cch remains in lieu of becoming çc; kṣ becomes kkh, not sk or ẖk; ṣṭ and ṣṭh give ṭṭh, not sṭ. In kiçça we have an older form than kīça, in ahakaṁ than ahake, hake, hage. In practically all these details we must see an earlier stage of what becomes Māgadhī in the grammarians. With it may be compared the metrical inscription of the Jogīmārā cave on the Rāmgarh hill which belongs to the period of Açoka. The Prākrit of the Gobaṁ° agrees with this Old Māgadhī in having l for r and e in the nominative singular, but it reduces all sibilants to s. It thus shows a certain similarity to the Ardha-Māgadhī of the grammarians, but that dialect often keeps r though it frequently alters it to l; for instance it has r for the kaleti of this Prākrit and the Old Māgadhī. Other points of similarity are the retention of the dental for cerebral in vanna; the lengthening of the vowel before the suffix ka (vannīkāhi); the accusative plural neuter in pupphā; and the infinitive bhuṁjitaye (bhuñjittae). There are points of difference, but they are probably all cases of earlier forms. Thus, as in Old Māgadhī, we have no softening or loss of intervocalic consonants; n is not cerebralised, but even introduced in palinata; ḷ appears in lieu of l; the instrumental in āhi has no nasal; the nominative of vat stems appears as in vā, as against vaṁ or vante; in the infinitive we find no doubling of the consonant in taye. The fact, however, of the regular change of r to l and the use of the form yeva after a long vowel as in Māgadhī and Pāli show that the Old Ardha-Māgadhī was more akin to Māgadhī than the later Ardha-Māgadhī, which came steadily under the influence of the western dialects as shown by the tendency to change e of the nominative to o.