form), or JihvamulTya {h) before the gutturals and Upadhmanlya (h) before the ^labials; e. g. indrah pdhca (i. 75). But as remains and ts us fs become is us rs': a. regularly in compounds in all the Samhitas; t. g. paras-pd- 'far-protecting'; havis-pa- 'drinking the offering'; ^«f-/5//- 'evil-doing'; dus-pdd- 'evil-footed'. The general rule, however, applies in the following compounds: purdh-prasravana- 'streaming forth'; chdndah-paksa- {KSf) 'borne on the wings of desire'; sreyah-keta- (AV.) 'striving after superiority'; sadyah-kri- (AV.) 'bought on the same day'; bahihparidhi (TS.) 'outside the enclosure'; itdh- pradana- (TS.) 'offering from hence (= this world)'.
The repeated (or amreiita) compounds also follow the general rule, doubtless from a desire to change the repeated word as httle as possible; thus ptirvah-purvo 'each first'; pardhparah 'always without' (AV.); pdrusah- parusas (VS.) 'from every knot'; pilrusah-puruso (TS.) 'every man'; pdrit/i- paruh (TS.) 'joint by joint', but pdrus-parur also in RV. AV. TS.
^. Often in external Sandhi in the RV.^; e.g. divds pdri (x. 45') 'from the sky'; pdtnivatas krdhi (i. 14^) 'make them possessed of wives'; dydus pita (iv. i '°) 'Father Heaven'.
d. Before mutes immediately followed by s or s, final s regularly becomes Visarjaniya; e. g. satdkratith tsdrat (viii. i"); ubhayatah-kmur (TS.) 'two-edged'. Occasionally the sibilant disappears, as in ddka ksdrantlr {wn. ^i^')i.
e. I. Before a simple sibilant final j' is either assimilated or becomes Visarjaniya; e.g. vas sivdtamo or vah Hvdtamo (x. 9'); devis sal or divih sal (x. 1285); nas sapdtna or nah sapdtna (x. 128'). Assimilation is undoubtedly the original Sandhi* and is required by some of the Pratisakhyas^; but the Mss. usually employ Visarjaniya, and European editions regularly follow this practice*
a. The sibilant disappears in the compounds barhi-sdd- 'sitting on the sacrificial litter'; dyau-samsita- (AV.) 'sky-sharpened'; and, after lengthening the preceding a, ayi- saya raja-saya hard-saya (TS. I. 2. 112 = MS. I. 27) for ayai-, rajas-, haras-.
2. Before a sibilant immediately followed by a voiceless mute, a final sibilant is dropped; e. g. mandibhi stomebhir (i. 9 3) for mand'bhis; mitJia- spfdhya (i. 16 69) for mithas-; du-siutl- 'ill praise' for dus-T . The omission is required by the Pratisakhyas of the RV.^ VS., TS., and is the practice of all the Mss. of the MS.
3. Before a sibilant immediately followed by a nasal or semivowel, a final sibilant is optionally dropped; thus kṛta śrávaḥ (vi. 5 8 3), beside which (though the Pp. reads ir/fl) the MS. reads kṛtah srdvah<r8>; ni-svardm (vii. i?) for nis-svardm 'noiseless' (Pp.^ however, ni-svardm).
79. Euphonic combination of final r. — As A is the pause form of both r and s, a certain amount of mutual contamination appears in their Sandhi; r, however, suffers much more in this respect than s. Since both j- and r when preceded hy t u have the same natural Sandhi, it is in a few
1 This treatment of final s before voice- less gutturals and labials, which is parallel to that before t, was doubtless the original from of sentence Sandhi.
2 ado pito (l. 1877) is probably only an apparent exception, as ado = Ada ?/, not adas (Pp. adah); the Paippalada recension of the AV., however, has adas fito, for adas.
3 Though the Pp. reads adha, the PB. in quoting the verse has adhah ; see Olden- BERG 369, note I.
4 Cp. Whitney on APr. 11. 40.
5 Cp. Wackernagel I, p. 342, top.
S u pdrTto sihcata (x. 1071), ito probably = ltd u (Pp. pdri iidh).
7 The omission was doubtless due to the fact that it made no difference to the pro- nunciation. Hence probably the wrong analysis of isastut by the Pp. as isak-stut, instead of isa-stut, as in isd-vant- (cp. BR.).
8 Cp. BOLLENSEN, ZDMG. 45, 24; PiSCHEL,
Vedische Studien I, 13.
