We must also determine whether a/kSitam is more archaic than a/kSiti from the standpoint of morphology. For this problem, I propose to examine the interrelation of the Indo-European formants *-to- and *-ti-, as reflected in the -ta- and -ti- of the adjectives a/kSita- and a/kSiti-. From the Indo-European point of view, as I argue elsewhere, adjectives in *-to- produce abstract substantives in *-ti-.1Accordingly, we expect an abstract substantive -kSiti/- on the basis of a verbal adjective -kSita/-. Although such a substantive is not directly attested in the Rig-Veda itself, there is no reason to doubt its archaism and we may safely derive the Rig-Vedic adjective a/kSiti- as a BahuvrIhi built from it, meaning 'who/which has no kSiti-'. The substantive is attested from the Atharva-Veda onward, with the accentuation (kSi/ti-) characteristic of compound ti- substantives which had switched to the status of non-compounds. There are in the Rig-Veda some rare negative abstract substantives in a/-...-ti-, apparently generated by the corresponding adjective in a/-...-ta-:
| a/-jI-ti- | 'absence of injury' |
| a/-juS-ti- | 'unpleasantness' (attested along with a/-juS-ta- 'unpleasant') |
| a/-di-ti- | 'boundlessness' (hence the goddess Aditi)2 |
s/ra/vas + amR/tam/a/MrtyuThen too, the privative verbal adjectives of mR- and kSi- are correlates elsewhere as well. For example, in the Rig-Vedic hymn 9.113 (stanza 7), the loka/-, or abode of the sun's eternal light, is described as both amR/ta- and a/kSita-:
s/ra/vas + a/kSitam/a/kSiti
amR/te loke/ a/kSiteTo sum up: by using the comparative method, it is possible to show that (s/ra/vas) a/kSitam is the basic variant conforming directly to Indo-European morphology, while a/kSiti (s/ra/vas) is the derivative variant, generated from within Indic morphology. Furthermore, while a/kSiti- is formed by genuine Indic grammatical rules, its lack of a solidly-precedented grammatical constituency suggests that it is an ad hoc poetic creation.(There is an interesting parallel in Greek. Just as the adjective a/kSiti- is built from an abstract substantive kSiti-, so also the Greek adjective ἀθάνατος 'immortal' seems to be built from the substantive θάνατος 'death' rather than from the adjective θνητός 'mortal'. The analogue to a/kSita- would have been *ἄθνητος, a hypothetical form which Kuryłowicz argues to be more archaic than the actual ἀθάνατος.5 The theoretical possibility of a form like *ἄθνητος is evident from such formal pairs as Homeric τλητός/ἄτλητος).Word-order is likewise a criterion for testing the relation of s/ra/va(s) a/kSitam to a/kSiti s/ra/vas. Since poetic form may blur current phases in the natural language, we may best ascertain the Indic trends in word-order from prose, such as that of the BrAhmaNas. There the basic positional behavior of attributes, either adjectives or genitives, is that they precede the substantive.6 Since the expression s/ra/va(s) a/kSitam does not conform to this prevailing Indic trend, we may infer that a/kSiti s/ra/vas is the more normal expression, from the standpoint of nonpoetic language. From the standpoint of poetic language, on the other hand, the adjective a/kSita- is more normal in that it seems more traditional, as I will try to show.The absence of the adjective a/kSiti- from collocations with any substantive other than s/ra/vas is the first indication that a/kSiti s/ra/vas may be secondary to and based on s/ra/va(s) a/kSitam. The intricate positional constraints on the distribution of the adjective a/kSita- in collocation with substantives other than s/ra/vas are an even more important positive indication that a/kSita- is a highly archaic component of traditional Rig-Vedic phraseology. Of the 11 occurrences of a/kSita- with neuter substantives, 9 follow the substantive rather than precede it. Furthermore, 8 times out of these 9 a/kSita- is in absolute verse-final position. This positioning within the closing shows the archaism of phrases containing a/kSita-, simply because the rhythmical inflexibility of the closing makes it harder to devise new phrases to fit the meter and easier to retain old phrases ready-made to fit the meter.Besides the 11 occurrences of a/kSita- with neuter substantives, this adjective is attested only with masculine substantives in the accusative (7 times) and locative (once), thus only in slots where the required masculine form of a/kSita- is not formally distinct from the neuter. Of the 8 masculine occurrences, 7 follow the substantive rather than precede it, and all 8 are in absolute verse-final position.Where a/kSita- precedes rather than follows its substantive, there seem to be concomitant requirements: (1) adjective and substantive are in different verses, as in a/kSitam#i/ndum of RV 9.26.2; or (2) adjective and substantive are in verse-medial position, as in #1 2 3 4 a/kSitam pA/yas 10 11 12# of 9.68.3; or (3) adjective and substantive are separated from each other by other lexical elements, as in #1 2 3 4 a/kSitam pA/yasA ra/ja# of 1.58.5. Included in the third category is a/-pra-kSitam va/su of 1.55.8. If we otherwise compare the 5 Rig-Vedic instances of a/- plus preverb plus -kSita- to the 19 of a/kSita-, the idiosyncratic patterns of the former category emerge: a/prakSita- (1.55.8) and a/vikSita- (7.1.24) occur in absolute verse-initial position, and the latter of the two is in a distinctly non-neuter form (nominative plural masculine in -Asas); but the other 4 out of the 5 instances are either masculine singular accusative or neuter singular, that is, in the same configuration as the attestations of a/kSita- without preverb-infixation. Of the three instances where a/- plus preverb plus -kSita- follows rather than precedes its substantive, all are in the archaic verse-final position.7In conclusion, both metrical and grammatical consideration lead me to propose that a/kSiti s/ra/vas had replaced an earlier s/ra/va(s) a/kSitam as a phrase suitable for ending a Rig-Vedic octosyllable. For the purpose of reinforcing this proposal, more work is required on the overall behavior of phraseology involving s/ra/vas and on the interaction of this behavior with that of Rig-Vedic meter. Before I can undertake such a task, however, I must attempt to present an overall picture of the basic Indic meters and their possible origins. For the moment, then, let us view the meters alone, without regard to the phraseology.
