before which ā is made long: this ending is found in no other noun-declension, and elsewhere only in the personal pronouns (of all numbers).
f. The gen. has sya added to the final a; and this ending is also limited to a-stems (with the single exception of the pronoun amúṣya: 501). Its final a is in only three cases made long in the Veda; and its y is vocalized (asia) almost as rarely.
g. The loc. ends in e (as if by combining the normal ending i with the final of the stem), without exception.
h. The voc. is the bare stem.
328. Dual. a. The dual endings in general are the normal ones.
b. The nom., acc., and voc. masc. end in the later language always in āu. In the Veda, however, the usual ending is simple ā (in RV., in seven eights of the occurrences). The same cases in the neut. end in e, which appears to be the result of fusion of the stem-final with the normal ending ī.
c. The instr., dat., and abl. have bhyām (in only one or two Vedic instances resolved into bhiām), with the stem-final lengthened to ā before it.
d. The gen. and loc. have a y inserted after the stem-final before os (or as if the a had been changed to e). In one or two (doubtful) Vedic instances (as also in the pronominal forms enos and yos), os is substituted for the final a.
329. Plural. a. The nom. masc. has in the later language the normal ending as combined with the final a to ās. But in the Veda the ending āsas instead is frequent (one third of the occurrences in RV., but only one twenty-fifth in the peculiar parts of AV.).
b. The acc. masc. ends in ān (for earlier āns, of which abundant traces are left in the Veda, and, under the disguise of apparent euphonic combination, even in the later language: see above, 208 ff.).
c. The nom. and acc. neut. have in the later language always the ending āni (like the an-stems: see 421; or else with n, as in the gen. pl., before normal i). But in the Veda this ending alternates with simple ā (which in RV. is to āni as three to two, in point of frequency; in AV., as three to four).
d. The instr. ends later always in āis; but in the Veda is found abundantly the more normal form ebhis (in RV., nearly as frequently as āis; in AV., only one fifth as frequent).
e. The dat. and abl. have bhyas as ending, with e instead of the final a before it (as in the Vedic instr. ebhis, the loc. pl., the gen. loc. du. [?], and the instr. sing.). The resolution into ebhias is not infrequent in the Veda.
f. The gen. ends in ānām, the final a being lengthened and having n inserted before the normal ending. The ā of the ending is not seldom (in less than half the instances) to be read as two syllables, aam: opinions are divided as to whether the resolution is historical or metrical only. A
