In the case of conjunct consonants we find that jñ gives in Bhāsa either ññ or ṇṇ, possibly the latter by error; Açvaghoṣa has ññ only, Kālidāsa ṇṇ. For ny and ṇy Bhāsa has always ṇṇ as against Açvaghoṣa’s ññ. The eliding of a consonant, with the compensatory lengthening of the vowel as in dīsadi, is unknown to Açvaghoṣa, where the omission of the consonant twice occurs but without lengthening; it is frequent in Bhāsa and regular in Kālidāsa. The analogous use of a short vowel and a double consonant to represent a long vowel with a single consonant is unknown to Açvaghoṣa, but Bhāsa has it in evva, evvaṁ, jovvana, devva, ekka. On the other hand, like Açvaghoṣa, for ry he has yy only in lieu of Kālidāsa’s jj. For the later metta matta is always found, and the epenthetic vowel is u, not i, in purusa, and puruva is normal. In inflection we have, in the nominative and accusative plural of neuter stems in a, āni in Açvaghoṣa, āṇi in Bhāsa, while both āṇi and āiṁ are allowed later. The accusative plural masculine has also, analogously to āni in the Ardha-Māgadhī of the Açoka inscriptions, [238] āṇi, and the locative singular feminine is in āaṁ, not as later āe. For the later attāṇaaṁ we have attāṇaṁ. For ‘we’ Açvaghoṣa has vayaṁ, Kālidāsa amhe; Bhāsa both and vaaṁ. In the genitive plural Bhāsa has both amhāaṁ and the only form later amhāṇaṁ, while Açvaghoṣa would doubtless have used amhākaṁ. kissa is kept for later kīsa, and kocci (kaccid) disappears later. The root darç is represented by dass and daṅs, grah by gaṇhadi against the later geṇhadi, which, however, is found in Açvaghoṣa. The older forms karia and gacchia or gamia, are found in lieu of kadua and gadua, but the last occurs once. mā is used with the gerund in the sense of alam. Many of these peculiarities mark also the Māgadhī, which appears in two slightly varied forms, the first in the Pratijñāyaugandharāyaṇa and the Cārudatta, the second in the Bālacarita and the Pañcarātra; in the two latter we have ṣ and o for the ç and e of the former. As in Açvaghoṣa there is no trace of obedience to the rules of the grammarians which require sṭ for Sanskrit ṣṭh or ṣṭ, çc for cch, sk or ẖk for kṣ. For ‘I’ we find ahake, which is an intermediate stage between Açvaghoṣa’s ahakaṁ and the later hage. ny becomes ṇṇ, not ññ, and the use of y to denote a dropped consonant is not carried out. The only passages that can claim to be anything like Ardha-Māgadhī are the remarks of Indra in disguise in the Karṇabhāra, where the characteristic signs, the use of r, s, and e, are found; in the speeches of Muṣṭika and Cāṇūra in the Bālacarita we have the use of l and a locative in ammi. A single passage in the Pañcarātra suggests Māgadhī Apabhraṅça, but is probably corrupt. 7. THE METRES OF THE DRAMAS