The formation of gerunds in Indic, it is well known, was sharply dichotomized: The suffixation applied to simplexes differed totally from that applied to compound verbs, i.e. verb stems preceded by adverbal particles. Thus we have a systematically suppletive inflexional paradigm. This synchronic Indic suppletion derived in turn, we shall see, from an Indo-European derivational suppletion; and the suppletion which resulted in the system of gerund formation formed only a part of a much more pervasive system of Indo-European derivational deverbal suppletion.
(1) The suffix added to simplexes was either -tva or -tv{ or -tl~aya; these seem to have been in a high degree interchangeable, and this state of affairs in the RV points to their partly moribund status whereby their earlier distinctive values were in tile process of being merged.
The nearly exclusively RV suffix -tv{ is exemplified by k!'-tv{ 'having made', ga-tv[ 'having gone', hi-tv[ 'having abandoned' (~/hd-), bha-tv{ 'having become', and ]ani-tv{ 'having produced'. Notice that, with the exception of]anitv{, the base in these formations is in zero-grade. We shall return to fanitvt below. The largely RV suffix -tvdya is exemplified by ga-tl,dya 'having gone', hi-tvd),a 'having abandoned', ha-tvaya 'having slain', yuk-tvaya 'having yoked', and drs-t, vaya 'having seen'. Once again, we see that the base is regularly in zero-grade. Both of these suffixes were alternants which were in a moribund phase s at the time of RV.
The most alive and productive alternant was surely -tva, wtfich is much better attested in AV than in RV. For this w'e may cite in compressed form: RV hi-tva 'having abandoned', ha-tva 'having slain', bha-tva-'having become', pF-tva 'having drunk', #u-tva 'having heard', yuk-tva 'having yoked'; AV tTr-tva 'having crossed', baddhva-'having bound'. Again we find the base in zero-gradc.
(2) The fact that all these formations require the base ha zero-grade makes it strictly undesirable to associate their history with a verbal noun in -tu-, as Macdonell (A Vedic Grammar for Students, Oxford 1916, p. 188) does, since such a nominalization, it is well known (cf. Macdonell pp. 192, 194, Meillet