________________
VARIANT ENDINGS -U,-AÜ AND -A IN THE APABHRAMŚA VERSES
21
distinct feature of the dialect of Hemacandra's verses, mainly Dohās, is the occurrence of the patterns 2 and 3 side by side in one and the same verse. As instances may be quoted 340, 2: garuā bharu pekkhevi (...) hai ki na juttaü, 343, 1: unhaü (...) sialu (...) sīalā, 343, 2: vippiaāraü (...) daddhā, 350, 2: appanā (...) appanaü, 352: piu ditthaü (...) addhā valaā mahihi gaa, 358, 2: jīviu kāsu na vallahaü (...) na itthu, and 445, 2: abbhā laggā (...) pahiu radantaü jāi.
As already indicated, Alsdorf believes the three endings involved to be related as follows: -aü (2) would be a rhythmically conditioned variant of -u (1), while -ā (3) would have been derived directly from -ai (2). However, on closer consideration this picture may have to be modified on several points.
To begin with, the conclusion that the alternation of -u and -aü was determined by a rhythmic factor seems to have been based on incomplete quantitative data. Alsdorf notes that 608 of the 801 instances of -aü occur after a heavy syllable. This means that still no less than 193 instances (approximately 25%) are found after a light syllable. More importantly, however, Alsdorf does not provide any quantitative information concerning the instances of -u in past participles with a heavy penultimate syllable (e.g. païtthu and itthu in 81, 4, 11, vuttu in 83, 8, 15, ditthu and ghittu in 83, 15, 11, and uttu in 83, 17, 11). The situation is thus far more complex than that of light penult + u (?) alternating with heavy penult + aü (608). The complete picture is the following: light penult + u (?) beside light penult + aü (193), and heavy penult + u (?) beside heavy penult + aü (608). With this, in combination with the unlikely fact that the supposed rhythmic factor would be operative only in certain word-classes, little is left of the usefulness of this factor in explaining the occurrence of the extended ending - as well as, it seems, the long ending - beside the short one.
If -aü was not a rhythmically conditioned alternant of -u, we will have to treat 1 and 2 as completely autonomous patterns. This opens the way for an alternative grouping of the three patterns. The basic dividing line seems to run between 1, on the one hand, and 2 and 3, on the other. We would have to do with two altogether different types of realization of the concord relation between head noun adjective and subject-nominal predicate: while in 2 and 3 the adjective and nominal predicate are marked, in 1 they are not. In this respect, patterns 2 and 3 together seem to anticipate the situation in New Indo-Aryan.56 While the long ending -ā of pattern 3 is as such actually found in New Indo-Aryan (see note 56), pattern 2 seems to occupy an intermediate position, in that it makes use of material inherited from Skt (-ka). Pattern 1, on the