After this chapter you will be able to join two Sanskrit words, or a stem and an affix, when the first ends in a vowel and the second begins with one. You will know the seven things that can happen at such a junction — a vowel turns into a semivowel (sudhy upāsyaḥ), e or o turns into ay or av (haraye), two vowels merge into one guṇa vowel (upendraḥ), into one vṛddhi vowel (kṛṣṇaikatvam), into the second vowel alone (prejate), into one long vowel (daityāriḥ), or the first vowel alone survives (hare 'va) — and which rule of the Siddhānta-kaumudī orders each. You will also meet the small machinery that rides along with vowel-sandhi: the optional doubling of consonants, the rule that a substitute is treated like what it replaced, and the name lopa for a sound that disappears. Each of the forty-three rules below is shown at work on a concrete word.
have → what this rule does → get. A form followed by a space and a second word (sudhy upāsyaḥ) is two padas in close utterance; a hyphen (gav-agram) marks a stage that is not yet a finished form.sudhī + upāsyaḥ → suddhy upāsyaḥWhen a word ending in ī meets a word beginning with u, the ī becomes y: sudhī upāsyaḥ becomes sudhy upāsyaḥ. That is the whole of vowel-to-semivowel sandhi (SK 47). But the new y now sits after the consonant dh, and Pāṇini's rules allow a consonant in that position to be doubled. Working out whether the doubling really applies — since the y was a vowel a moment ago — takes six rules (SK 48–52 and SK 54), and Bhaṭṭoji uses the occasion to teach the general principle that a substitute is treated like its original (SK 49–51), the name lopa (SK 53), and the other doubling rules of the grammar (SK 55–60). The finished running example, with the doubling taken, is suddhy upāsyaḥ.
Rule. In place of an ik vowel (i u ṛ ḷ, long or short) comes the matching yaṇ (y v r l) when a vowel follows, in close utterance. "Matching" is decided by nearness of place: i and y are both palatal, u and v both labial, ṛ and r both made at the roof of the mouth, ḷ and l both dental.
Example.
sudhī upāsyaḥ → the ī of sudhī is an ik; the u of upāsyaḥ is a vowel after it → the ī is replaced by the palatal yaṇ, y → sudhy upāsyaḥ.
Applies when / blocked by. Only within saṃhitā, the close utterance of chapter 1 (SK 28): the word saṃhitāyām is carried into this rule from an earlier heading. Scholars discuss exactly how the sūtra's wording secures that the vowel must follow immediately.
Watch out. The rule replaces the ī only; nothing happens to the u. Compare Section C, where one vowel replaces both.
Rule. A yar consonant standing after a vowel may optionally be doubled, when what follows it is not a vowel. Yar is the pratyāhāra from y to the marker r of the Śivasūtras: every consonant except h. The word "two" (dve) is carried in from SK 59, which stands just before this rule in Pāṇini's order; "optionally" (vā) comes from the rule before that (A 8.4.45), which is not part of this chapter.
Example.
sudhy upāsyaḥ → the dh stands after the vowel u and before y, which is not a vowel → the dh may be doubled → sudhdhy upāsyaḥ (the cluster is tidied by SK 52 below).
Applies when / blocked by. Only when no vowel follows the consonant. Here is the difficulty: the y after dh used to be the vowel ī. If the y still counts as ī, the dh stands before a vowel and cannot double. The next three rules settle this.
Rule. A substitute is treated like its original (it is sthānivat, "like the sthānin"), except in an al-vidhi: a rule whose working depends on the actual sound-form of the original. Al is the pratyāhāra of all sounds; an al-vidhi is a rule keyed to a particular sound as such.
Example.
sudhy upāsyaḥ → is the y to be treated as the ī it replaced, so that dh is "before a vowel"? → the doubling rule SK 48 looks at the sound after dh, so it is an al-vidhi; there the substitute is not treated like its original → the y is simply y, and the doubling is not blocked.
Watch out. Being sthānivat is a general convenience of the grammar: a substitute inherits the names and roles of what it replaced. This rule gives the convenience and, in the same breath, its main limit.
Rule. A substitute for a vowel, when the substitution was caused by what follows (it is para-nimitta), is treated like its original even in an al-vidhi, provided the operation is to be done on what precedes the original vowel. Bhaṭṭoji says plainly that this rule exists for the sake of al-vidhis.
Example.
sudhy upāsyaḥ → the y replaced ī because the vowel u followed, so it is a para-nimitta vowel-substitute → the doubling of SK 48 would be done on the dh, which precedes the ī → by this rule the y would again count as the vowel ī, and the doubling would again be blocked.
Applies when / blocked by. This rule re-opens the difficulty that SK 49 had closed. It is deliberately broad; SK 51 lists the operations in which it does not hold.
Rule. The treatment-as-original of SK 50 is switched off for ten kinds of operation: those concerning the end of a pada (padānta), doubling (dvirvacana), the affix vara, the dropping of y (ya-lopa), accent (svara), a savarṇa substitute, anusvāra, lengthening (dīrgha), a jaś substitute and a car substitute.
Example.
sudhy upāsyaḥ → the operation in question is doubling → doubling is on the list, so the y is not treated as ī → the dh stands before a non-vowel and SK 48 may double it → sudhdhy upāsyaḥ.
Applies when / blocked by. SK 67 will use the first item of the list, padānta, to stop a wrong semivowel in kāni santi.
Rule. In place of a jhal consonant comes the matching jaś consonant when a jhaś consonant follows. Jhal is every stop and sibilant together with h (from jha to the marker l); jhaś is the ten voiced stops (jha to the marker ś); jaś is the five unaspirated voiced stops j b g ḍ d. In plain words: before a voiced stop, a stop loses its aspiration and becomes voiced. Bhaṭṭoji calls the rule self-evident.
Example.
sudhdhy upāsyaḥ → the first dh is a jhal and the second dh after it is a jhaś → the first dh becomes the jaś of its own class, d → suddhy upāsyaḥ, the running form.
Rule. This is a saṃjñā, a naming rule. When a sound that had occasion to appear does not appear, that non-appearance is named lopa (disappearance). A rule that orders lopa orders the sound to vanish.
Example.
sudhy → the next rule threatens to order lopa of the y → if it did, the y would simply not be heard → sudh (this is exactly what the following rules prevent).
Watch out. The name is given to the non-appearance of something that was due (prasakta), not to the mere absence of a sound. The word matters in SK 54, 60, 62, 67 and 71.
Rule. When a pada ends in a consonant cluster, its final sound disappears (lopa). The word "of a pada" is carried in from an earlier heading, and only the last sound goes.
Example.
suddhy upāsyaḥ → suddhy is a pada ending in the cluster ddhy → its final y is due to disappear → suddh upāsyaḥ (wrong).
Applies when / blocked by. Two vārttikas rescue the y. The first, yaṇaḥ pratiṣedho vācyaḥ, says the disappearance must be prohibited for a yaṇ: so the y stays. The second, yaṇo mayo dve vācye, orders a doubling that involves a yaṇ and a may. May is the pratyāhāra from ma to the marker y: the four nasals m ṅ ṇ n and the twenty other stops, from jh to p — twenty-four sounds in all. Read one way (mayaḥ "after a may", yaṇaḥ "of a yaṇ"), the vārttika doubles a yaṇ that follows a may — so the y after dh may double as well. Bhaṭṭoji's commentators take both readings to be intended.
Watch out. With the dh optionally doubled (SK 48) and the y optionally doubled (the vārttika), the running example has four correct forms: one dh and one y; two of each; two dh and one y; one dh and two y. Bhaṭṭoji names one word for each of the four yaṇ sounds: suddhyupāsyaḥ (y), maddhvariḥ (madhu + ariḥ, the enemy of the demon Madhu, v), dhātraṃśaḥ (r), lākṛtiḥ (ḷ + ākṛtiḥ, "shaped like the sound ḷ", l). Bhaṭṭoji's commentators note that the first vārttika is rejected in the Mahābhāṣya.
Rule. The t of the word putra is not doubled before ādinī ("eater") when abuse (ākrośa) is meant.
Example.
putra + ādinī → SK 48 would double the t (it stands after u and before r) → in an insult the doubling is refused → putrādinī tvam asi pāpe ("you are a son-eater, wicked woman").
putra + ādinī, said as plain fact → the doubling goes ahead → puttrādinī sarpiṇī ("a she-snake that eats her young").
Applies when / blocked by. A vārttika, tatpare ca, extends the refusal to the case where ādinī follows a second putra: putraputrādinī tvam asi pāpe ("eater of a son's sons"). Another, vā hatajagdhayoḥ, makes the doubling optional before hatī and jagdhī: putrahatī or puttrahatī, putrajagdhī or puttrajagdhī.
Rule. When three or more consonants are joined together, the doubling is optional in the opinion of the teacher Śākaṭāyana.
Example.
indraḥ → the cluster ndr has three consonants → the n may be doubled or not → inndraḥ or indraḥ.
rāṣṭram → the cluster ṣṭr has three consonants → rāṣṣṭram or rāṣṭram.
Rule. In the opinion of Śākalya there is no doubling at all, in any position.
Example.
arkaḥ → the k after r could double by SK 59 → in Śākalya's view it does not → arkaḥ.
brahmā → likewise no doubling → brahmā.
Watch out. Naming a teacher's opinion is Pāṇini's way of stating an option: the forms with and without doubling are both correct.
Rule. After a long vowel there is no doubling, in the opinion of the teachers (ācāryas).
Example.
dātram → the t stands after the long ā → no doubling → dātram.
pātram → likewise → pātram.
Rule. A yar consonant standing after r or h, when that r or h itself stands after a vowel, is optionally doubled. This is the rule from which SK 48 borrows its "two"; the "optionally" of both rules is carried in from the rule before this one (A 8.4.45).
Example.
hari + anubhavaḥ → SK 47 gives hary anubhavaḥ → the y stands after r, which stands after a → the y may be doubled → haryyanubhavaḥ or haryanubhavaḥ.
na hi + asti → nahy asti → the y stands after h, which stands after i → nahyyasti or nahyasti.
Rule. A yam sound standing after a consonant optionally disappears when the same yam sound follows. Yam is the pratyāhāra from ya to the marker m: the semivowels y v r l and the nasals ñ m ṅ ṇ n. "The same" is secured by reading the sūtra's "of yams … before a yam" in matching order: y drops before y, m before m, and so on.
Example.
haryyanubhavaḥ → the first y stands after the consonant r and before another y → it may disappear → haryanubhavaḥ.
ādity + ya + m (the word ādityam built from āditya with an affix ya, the a of āditya having been dropped by a later rule) → the y after t stands before y → it may disappear → ādityam (the received form).
Applies when / blocked by. In haryanubhavaḥ the result is the same as simply not doubling, so Bhaṭṭoji points to ādityam as the case where this rule alone gives the received form. It does not touch māhātmyam: there m stands before y, and the rule drops a yam only before the same yam.
hare + e → haraye (also viṣṇo + e → viṣṇave)Pāṇini gives the ec vowels a different substitute from the one the ik vowels took in Section A. Before a vowel, each of the four is replaced by a vowel-plus-semivowel pair: e → ay, o → av, ai → āy, au → āv. The dative haraye ("to Hari") shows it: at the stage hare + e, the e of hare becomes ay. The group also covers a special av before y-affixes (SK 63–64), three fixed forms in -ayya (SK 65–66), and the optional dropping of the new y or v at the end of a word (SK 67).
Rule. In place of an ec vowel (e o ai au) come ay, av, āy, āv respectively, when a vowel follows. "Respectively" (yathāsaṃkhya) means item to item in order: the first for the first, the second for the second, and so on.
Example.
hare + e (the stem hari has already become hare before the dative ending e) → the e of hare is an ec before a vowel → it is replaced by the first substitute, ay → haray + e → haraye.
viṣṇo + e → the o is the second ec → av → viṣṇave.
nai + akaḥ → ai → āy → nāyakaḥ; pau + akaḥ → au → āv → pāvakaḥ.
Watch out. Scholars discuss whether SK 47 and this rule could be written as one; that is a point about economy, not about the forms.
Rule. A marker (it) disappears: "of that (the it) there is lopa." The rule states what happens to every marker found by the rules of chapter 1.
Example.
haray + e → the y of ay is a final consonant in the statement of a substitute, so by chapter 1 it might be taken for a marker and dropped → but the y and v are pronounced for a purpose (uccāraṇa-sāmarthya): without them the substitutes would be mere a and ā → they are not markers and do not drop → haraye, viṣṇave, nāyakaḥ, pāvakaḥ.
Watch out. The argument from "pronounced for a purpose" cancels both the lopa and the marker-name itself: the y of ay is not an it at all.
Rule. Before an affix beginning with y, in place of o and au come the v-ending substitutes (vānta) av and āv. SK 61 needs a following vowel; here what follows is the consonant y, so this separate rule is needed.
Example.
go + ya (the affix yat, "product of", added by a later rule) → the o stands before a y-affix → av → gavyam ("a product of the cow").
nau + ya (the affix yat, "to be crossed by") → au → āv → nāvyam ("to be crossed by boat").
Applies when / blocked by. A vārttika, gor yūtau chandasi, adds go before yūti in the Veda, and a second, adhvaparimāṇe ca, adds it when a measure of distance is meant: go + yūtiḥ → gavyūtiḥ. Bhaṭṭoji adds a remark on how the sūtra's wording secures that the v is really heard.
Rule. When the o or au belongs to a dhātu, the v-ending substitute of SK 63 applies only to an o or au that was itself produced because of that very y-affix, not to any other.
Example.
lo + ya (the root lū, its ū raised to o because the affix yat follows) → the o is due to the y-affix → av → lavyam.
avaśya-lau + ya → likewise, au due to the affix → āv → avaśyalāvyam.
oyate, auyata → here the o and au are not produced by the y-affix → no av, āv → the forms stay as they are.
Rule. The forms kṣayya and jayya are fixed as correct (nipātana, fixing a form by simply stating it) in the sense "capable of being destroyed / conquered". What is fixed is the y-ending substitute ay in place of the e these words would otherwise show.
Example.
kṣe + ya ("able to be destroyed") → the e stands before the consonant y, so SK 61 gives nothing → this rule fixes ay → kṣayyam.
je + ya ("able to be conquered") → jayyam.
Applies when / blocked by. Only in the sense "capable of". In the sense "fit to be, worthy of", the ordinary forms stand: kṣeyaṃ pāpam ("sin fit to be destroyed"), jeyaṃ manaḥ ("a mind fit to be conquered").
Rule. The form krayya is fixed (nipātana) in the sense "meant for that", that is, meant for buying.
Example.
kre + ya → in the sense "displayed in the market so that buyers may buy" → ay is fixed → krayyam.
kre + ya → in the sense "worth buying" → the ordinary form → kreyam.
Rule. In the opinion of Śākalya — that is, optionally — a y or v at the end of a pada, standing after a or ā, disappears when an aś sound follows. Aś is the pratyāhāra from a to the marker ś: all vowels and all voiced consonants.
Example.
hare + ehi → SK 61 → haray ehi → the y ends a pada, stands after a, and e follows → the y may disappear → hara ehi or harayehi.
viṣṇo + iha → viṣṇav iha → viṣṇa iha or viṣṇaviha.
śriyai + udyataḥ → śriyāy udyataḥ → śriyā udyataḥ or śriyāyudyataḥ.
gurau + utkaḥ → gurāv utkaḥ → gurā utkaḥ or gurāvutkaḥ.
Applies when / blocked by. Once the y is gone, a stands before e — but no further vowel-sandhi happens. This rule belongs to the Tripādī and is asiddha (chapter 1, SK 12): for the sandhi rules of Sections C–G the y counts as still there.
Watch out. In kāni santi and kau staḥ, the s of santi and staḥ is the root as with its a dropped. If that dropped a were still counted present (SK 50), the i of kāni would become y and the au of kau would become āv. SK 51 forbids the treatment-as-original in an operation at the end of a pada, so kāni and kau stand.
upa + indraḥ → upendraḥ (also gaṅgā + udakam → gaṅgodakam, kṛṣṇa + ṛddhiḥ → kṛṣṇarddhiḥ)From here to the end of the chapter, one vowel replaces two. When a or ā meets i, u or ṛ, the pair is replaced by the guṇa vowel nearest to the second: a + i → e (upendraḥ), ā + u → o (gaṅgodakam), a + ṛ → ar (kṛṣṇarddhiḥ). The heading rule SK 68 says that the substitute stands for both vowels; SK 69 orders the guṇa; SK 70 explains why the guṇa of ṛ carries an r with it; SK 71 tidies the consonant cluster that results.
Rule. This is an adhikāra, a heading whose words are read into every rule that follows it through SK 89: the single substitute ordered stands in place of both the preceding sound and the following one. Such a substitute is called an ekādeśa, "one substitute (for two)".
Example.
upa + indraḥ → the following rule orders a guṇa → by this heading the guṇa replaces both the a and the i, not one of them → upendraḥ.
Rule. After a or ā (the a-varṇa, "the a-sound" in all its kinds), when a vowel follows, a single guṇa vowel replaces both, in close utterance. Guṇa is a e o (chapter 1, SK 17); the one chosen is the nearest in place to the vowel replaced: e for i, o for u, a for ṛ.
Example.
upa + indraḥ → a before i → one guṇa for both; the palatal guṇa is e → upendraḥ.
ramā + īśaḥ → ā before ī → e → rameśaḥ.
gaṅgā + udakam → ā before u → the labial guṇa o → gaṅgodakam.
kṛṣṇa + ṛddhiḥ → a before ṛ → the guṇa nearest to ṛ is a → kṛṣṇa-ddhiḥ, completed by SK 70.
Applies when / blocked by. SK 72 (vṛddhi before e o ai au), SK 73–74 and SK 78 override it in their own cases. The word "in saṃhitā" is carried in, as in SK 47.
Rule. An aṇ vowel (a i u) that comes in place of ṛ always operates rapara, "with r after it". Bhaṭṭoji recalls from chapter 1 that ṛ is the name of thirty sounds — the eighteen kinds of ṛ and the twelve of ḷ, which are savarṇa — and that the "r" here is the pratyāhāra ra of chapter 1 (SK 2), which covers r and l. So the substitute for ṛ takes r, that for ḷ takes l, by nearness.
Example.
kṛṣṇa + ṛddhiḥ → SK 69 gives the guṇa a in place of a + ṛ → an aṇ replacing ṛ comes with r → kṛṣṇar + ddhiḥ → kṛṣṇarddhiḥ.
tava + ḷkāraḥ → guṇa a in place of a + ḷ → it comes with l → tavalkāraḥ.
Applies when / blocked by. In kṛṣṇarddhiḥ the d now stands after r, so it may be doubled by SK 59; the next rule sorts out the resulting forms.
Rule. A jhar consonant standing after a consonant optionally disappears when a savarṇa jhar follows. Jhar is the pratyāhāra from jha to the marker r: all stops and sibilants (not h). "Savarṇa" here means of the same class, as in chapter 1 (SK 10).
Example.
kṛṣṇar + ddhiḥ → the d stands after r and before dh, a stop of its own class → it may disappear; and by SK 59 it may also be doubled → three forms:
no doubling, d dropped → kṛṣṇardhiḥ;
no doubling and no dropping, or doubling and dropping → kṛṣṇarddhiḥ;
doubling and no dropping → kṛṣṇardddhiḥ (two d before the dh, three stops in all).
Watch out. Tavalkāraḥ has four forms, by a different pair of options: the l after a may double by SK 48, and the k after the yaṇ l may double by the vārttika of SK 54 (on the reading "a may after a yaṇ"). Bhaṭṭoji sums it up in a verse: the doubling of l alone, of k alone, of neither, or of both — tavalkāraḥ, tavallkāraḥ, tavalkkāraḥ, tavallkkāraḥ. In print the third form of kṛṣṇarddhiḥ differs from the second only by the length of its cluster.
kṛṣṇa + ekatvam → kṛṣṇaikatvam (also upa + eti → upaiti, upa + ṛcchati → upārcchati)When a or ā meets e o ai au, the guṇa of Section C is overridden and the two vowels become one vṛddhi vowel: a + e → ai (kṛṣṇaikatvam), ā + o → au (gaṅgaughaḥ). Two special rules extend vṛddhi to places where guṇa or the "second vowel wins" of Section E would otherwise apply: before the verbs eti and edhati and the substitute ūṭh (SK 73), and after a prefix before a root beginning with ṛ (SK 74, made optional for noun-derived roots by SK 77). SK 75–76 explain why the r of upārcchati does not turn into visarga.
Rule. After a or ā, when an ec vowel (e o ai au) follows, a single vṛddhi vowel replaces both. Vṛddhi is ā ai au (chapter 1, SK 16); the one chosen is the nearest: ai for e and ai, au for o and au. This is an apavāda, an exception, to the guṇa of SK 69: where both could apply, this one wins.
Example.
kṛṣṇa + ekatvam → a before e → one vṛddhi for both → kṛṣṇaikatvam.
gaṅgā + oghaḥ → ā before o → gaṅgaughaḥ.
deva + aiśvaryam → a before ai → devaiśvaryam.
kṛṣṇa + autkaṇṭhyam → a before au → kṛṣṇautkaṇṭhyam.
Rule. After a or ā, a single vṛddhi replaces both vowels before the verbs eti ("goes", the root iṇ) and edhati ("thrives", the root edh) when they begin with an ec, and before ūṭh (a substitute ū, written with its marker ṭh). It is an exception both to the "second vowel wins" of SK 78 and to the guṇa of SK 69.
Example.
upa + eti → a before the e of eti → vṛddhi → upaiti.
upa + edhate → upaidhate.
praṣṭha + ūh (the ū is ūṭh, which replaces the v of vāh by a later rule) → a before ū → vṛddhi au → praṣṭhauhaḥ.
Applies when / blocked by. Only when the two verbs begin with an ec. In upa + itaḥ → upetaḥ the root iṇ appears as i, and in pra + ididhat → mā bhavān predidhat ("may your honour not thrive") the root edh appears as idh: plain guṇa. Bhaṭṭoji adds that an exception stated earlier blocks only the rule that follows it next (purastād-apavāda-nyāya): this vṛddhi blocks SK 78 but not SK 80, so ava + ehi gives avehi by SK 80 and avaihi is wrong.
Watch out. Five vārttikas add vṛddhi in fixed words. akṣād ūhinyām: akṣa + ūhinī → akṣauhiṇī (an army; the n later becomes ṇ). svād īreriṇoḥ: sva + īrin → svairī ("one who moves by his own will"), and by the paribhāṣā on gender-marked forms the feminine svairiṇī. prād ūhoḍhoḍhyeṣaiṣyeṣu: after pra before ūha, ūḍha, ūḍhi, eṣa, eṣya → prauhaḥ, prauḍhaḥ, prauḍhiḥ, praiṣaḥ, praiṣyaḥ; but proḍhavān, because ūḍha names only the whole participle, not a part of ūḍhavān, and preṣaḥ, preṣyaḥ from the two roots īṣ with a long vowel, which take guṇa. ṛte ca tṛtīyāsamāse: sukhena ṛtaḥ → sukhārtaḥ in a compound with an instrumental, but paramartaḥ where the compound is of another kind. pravatsatarakambalavasanārṇadaśānām ṛṇe: before ṛṇa, after pra, vatsatara, kambala, vasana, ṛṇa, daśa → prārṇam, vatsatarārṇam, ṛṇārṇam (a debt taken to pay off a debt), daśārṇaḥ (a country; daśārṇā is a river, ṛṇa there meaning rough ground and water).
Rule. After an upasarga ending in a or ā, when a dhātu beginning with ṛ follows, a single vṛddhi replaces both vowels. The vṛddhi of ṛ is ā, and by SK 70 it comes with r: ār.
Example.
upa + ṛcchati → a before the ṛ of a root, after a prefix → vṛddhi ār for both → upārcchati.
pra + ṛcchati → prārcchati.
Applies when / blocked by. Without this rule SK 69 would give guṇa, uparcchati. The rule reads ṛti, with t after ṛ: by chapter 1 (SK 15) this names only the short ṛ.
Rule. The single substitute of SK 68 is treated as the final sound of the preceding element and as the initial sound of the following one (antavat, ādivat). So whatever a rule says about the end of a word applies to it, and whatever a rule says about the beginning of a word applies to it too.
Example.
upārcchati → the ār replaced the a of the pada upa and the ṛ of the root → as the final of upa, its r counts as the last sound of a pada → a rule about pada-final r now threatens it (SK 76).
Watch out. Bhaṭṭoji's commentators give the two sides of the rule in kṣīrapeṇa (the e counts as the final of kṣīrapa) and khaṭvā (the ā counts as the initial of the feminine affix). Only the roles of a group of sounds — being a pada, an affix, a stem — are extended, not the mere sound-value of one letter.
Rule. For an r at the end of a pada, visarga (ḥ) is substituted when a khar consonant follows or at pause. Khar is the pratyāhāra from kha to the marker r: the voiceless stops and the sibilants.
Example.
upārcchati → by SK 75 the r of ār is pada-final, and cch (a khar) follows → visarga is due → upāḥcchati (wrong).
Bhaṭṭoji's answer → an r that is pada-final only by the treatment-as-final of SK 75 does not take visarga → upārcchati. The proof is Pāṇini's own wording in rules such as ubhayatharkṣu (SK 3630) and carṣidevatayoḥ (SK 3167), where such an r stands unchanged.
Applies when / blocked by. Bhaṭṭoji uses the same passage to explain the word dhātau in SK 74. Since an upasarga already implies a root, the word is superfluous; so the sūtra is split in two (yoga-vibhāga) and dhātau alone re-orders the vṛddhi. The effect: the optional leaving-unchanged (prakṛti-bhāva) taught by ṛty akaḥ (SK 92, next chapter) does not apply here, and upārcchati is the only form.
Rule. After an upasarga ending in a or ā, before a sub-dhātu beginning with ṛ — a root made out of a noun, such as ṛṣabhīyati "wishes a bull for himself" — the vṛddhi of SK 74 is optional. Āpiśali is named as a courtesy, not as a condition.
Example.
pra + ṛṣabhīyati → a before the ṛ of a noun-derived root → vṛddhi or not → prārṣabhīyati or prarṣabhīyati (guṇa by SK 69).
pra + ḷkārīyati → ḷ is savarṇa with ṛ, so it counts too → prālkārīyati or pralkārīyati.
upa + ṝkārīyati → the rule names only the short ṛ (SK 74), so no vṛddhi → Bhaṭṭoji gives the two forms upaṛkārīyati and uparkārīyati (scholars discuss the first).
Watch out. Bhaṭṭoji's commentators give the ḷ and ṝ examples with slightly different prefixes; the forms above are Bhaṭṭoji's own.
pra + ejate → prejate (also upa + oṣati → upoṣati)After a prefix, a root beginning with e or o does not take the vṛddhi of Section D. Instead the two vowels are replaced by the second one alone: pra + ejate → prejate, upa + oṣati → upoṣati. Pāṇini calls this the pararūpa, "the form of the following". The group also teaches the name ṭi for the tail-end of a word (SK 79), because in a list of fixed words the pararūpa replaces that whole tail; and three rules on the imitation of noises (paṭat, paṭat paṭat) that use the pararūpa and introduce the name āmreḍita for a repeated word.
Rule. After an upasarga ending in a or ā, when a dhātu beginning with eṅ (e or o) follows, the single substitute for both vowels is the pararūpa: the following vowel's own form.
Example.
pra + ejate → a before the e of a root, after a prefix → the two are replaced by e → prejate (not praijate by SK 72).
upa + oṣati → a before o → replaced by o → upoṣati.
Applies when / blocked by. Bhaṭṭoji carries the words vā supi ("optionally, before a noun-derived root") of SK 77 into this rule by reading it as a second sentence (vākya-bheda). So before a noun-derived root beginning with e or o the pararūpa is optional: upa + eḍakīyati → upeḍakīyati or upaiḍakīyati; pra + oghīyati → proghīyati or praughīyati. A vārttika, eve cāniyoge, adds pararūpa before the particle eva when it does not mean restriction ("only"): kva + eva → kveva bhokṣyase ("where indeed will you eat?"); but tava + eva → tavaiva ("yours only"), where it does.
Rule. This is a saṃjñā. The part of a form that begins with its last vowel — that vowel and whatever consonants follow it — is named ṭi. In śaka the ṭi is a; in sīman it is an; in manas it is as.
Example.
śaka + andhuḥ → a vārttika, śakandhvādiṣu pararūpaṃ vācyam, orders pararūpa in the śakandhu group, and Bhaṭṭoji adds "and that, of the ṭi" → the ṭi of śaka and the following a are replaced by a → śakandhuḥ (not śakāndhuḥ by SK 85).
karka + andhuḥ → karkandhuḥ; kula + aṭā → kulaṭā.
sīman + antaḥ → the ṭi is an → the whole of it and the a of antaḥ give way to a → sīmantaḥ (the parting of the hair; sīmāntaḥ means something else).
manas + īṣā → the ṭi as and the ī give way to ī → manīṣā; likewise halīṣā, lāṅgalīṣā, patat + añjaliḥ → patañjaliḥ, sāra + aṅgaḥ → sāraṅgaḥ (a beast or bird; sārāṅgaḥ otherwise), mṛta + aṇḍaḥ → mṛtaṇḍaḥ, whence mārtaṇḍaḥ.
Applies when / blocked by. The group is an ākṛtigaṇa, an open list defined by its type: words like these are added from usage. A second vārttika, otvoṣṭhayoḥ samāse vā, makes pararūpa optional before otu (a cat) and oṣṭha (a lip) in a compound: sthūla + otuḥ → sthūlotuḥ or sthūlautuḥ; bimba + oṣṭhaḥ → bimboṣṭhaḥ or bimbauṣṭhaḥ; but tava + oṣṭhaḥ → tavauṣṭhaḥ, no compound.
Watch out. The pararūpa here replaces the ṭi, which can be more than one sound, not merely the last vowel. That is why the name is taught at this point.
Rule. After a or ā, when om or the particle āṅ (the prefix ā, cited with its marker ṅ) follows, the single substitute is the pararūpa.
Example.
śivāya + om → a before the o of om → replaced by o → śivāyoṃ namaḥ.
ā + ihi → SK 69 → ehi; then śiva + ehi → a before the e that began as āṅ → replaced by e → śivehi (not śivaihi by SK 72).
Applies when / blocked by. SK 73 does not block this rule (see there): avehi is right and avaihi wrong.
Rule. An avyakta-anukaraṇa is a word imitating an indistinct noise, such as a tree falling. When such a word stands before iti ("thus"), the a-sound inside it, together with the i of iti, is replaced by the single vowel i: the pararūpa.
Example.
paṭat + iti (the noise of falling, "thus") → the a of paṭat and the i of iti are replaced by the single vowel i → paṭiti (obligatory; paṭat iti is only the form before sandhi). How the t goes is not spelled out by Bhaṭṭoji.
Applies when / blocked by. A vārttika, ekāco na, exempts an imitation with a single vowel: śrat + iti → śrad iti, with no pararūpa (the d is by SK 84).
Watch out. Bhaṭṭoji's commentators point out that the rule must be read as "before the first i of iti", or the whole word iti would be replaced.
Rule. The pararūpa of SK 81 does not apply to an āmreḍita — the second member of a repeated word, named in SK 83. But for the final sound of the āmreḍita, provided it is just a t, the pararūpa is optional.
Example.
paṭat paṭat + iti → the second paṭat is an āmreḍita → its a is not merged with iti → but its final t and the i may be replaced by i → paṭatpaṭa-iti (finished at SK 83), or the t kept (SK 84).
Applies when / blocked by. A vārttika, ḍāci bahulaṃ dve bhavataḥ, licenses the repetition itself: before the affix ḍāc a word is doubled "variously", and that "variously" covers paṭat paṭat.
Rule. This is a saṃjñā. Of a doubled expression, the second member is named āmreḍita. The "of that" refers to the doubling ordered by the rule just before it in the Aṣṭādhyāyī.
Example.
paṭat paṭat + iti → the second paṭat is the āmreḍita → by SK 82 its t and the i of iti are replaced by i → paṭatpaṭa + iti → by SK 69 a + i → e → paṭatpaṭeti.
Watch out. Bhaṭṭoji's commentators state the step as "the pararūpa i in place of the t and the i, then guṇa"; the derivation is given here exactly as they give it.
Rule. At the end of a pada, a jhal consonant (SK 52) is replaced by the jaś of its class: in plain words, a jhal becomes the plain voiced stop of its class.
Example.
paṭatpaṭat + iti → on the option where the final t of the āmreḍita keeps its own form (SK 82) → the t ends a pada → it becomes d → paṭatpaṭad iti.
śrat + iti → śrad iti (SK 81).
Watch out. On this option the a of the second paṭat would merge with iti by SK 81 and give paṭatpaṭiti, were it not for the prohibition in SK 82. That prohibition is what keeps the āmreḍita whole.
daitya + ariḥ → daityāriḥ (also śrī + īśaḥ → śrīśaḥ, viṣṇu + udayaḥ → viṣṇūdayaḥ)When an ak vowel (a i u ṛ ḷ) meets a vowel of its own kind, the two become one long vowel: a + a → ā, ī + ī → ī, u + u → ū. One sūtra does it all (SK 85); two vārttikas settle what "long" means for ṛ and ḷ.
Rule. In place of an ak vowel and a following savarṇa vowel comes the single long vowel of that kind. Savarṇa is "of the same kind" as defined in chapter 1 (SK 10); long and short i, for instance, are savarṇa.
Example.
daitya + ariḥ → a before a → one long ā for both → daityāriḥ.
śrī + īśaḥ → ī before ī → śrīśaḥ.
viṣṇu + udayaḥ → u before u → viṣṇūdayaḥ.
Applies when / blocked by. The following sound must be a vowel: in kumārī śete the ī stands before ś, and nothing happens (Bhaṭṭoji discusses why the word ac is needed to secure this; scholars discuss his reasoning). The preceding sound must be an ak: in hare + e the e is an ec, so SK 61 gives haraye, not a long vowel. Bhaṭṭoji remarks that the sūtra could simply have said "an ak before an ak becomes long".
Watch out. Two vārttikas concern ṛ and ḷ. ṛti savarṇe ṛ vā: before a savarṇa ṛ, the substitute is optionally ṛ — hotṛ + ṛkāraḥ → hotṛkāraḥ or hotṝkāraḥ. ḷti savarṇe ḷ vā: before a savarṇa ḷ, optionally ḷ — hotṛ + ḷkāraḥ → hotḷkāraḥ or, since ṛ and ḷ are savarṇa, hotṝkāraḥ. Bhaṭṭoji explains that the substitute in each vārttika is a two-mātrā sound: two r (or two l) in the middle sharing one mātrā, and the vowel-parts on either side sharing the other. In both cases the next chapter's ṛty akaḥ (SK 92) will also allow the vowels to be left unchanged.
hare + ava → hare 'va; then go + agram, go + indraḥAt the end of a pada, e or o followed by a short a does not become ay or av. Instead the pair is replaced by the first vowel alone: hare + ava → hare 'va. Pāṇini calls this the pūrvarūpa, "the form of the preceding"; in writing, the sign ' (avagraha) marks where the a was. The word go ("cow") at the end of a pada has three rules of its own: it may be left unchanged before a (SK 87), its o may become av before any vowel (SK 88), and must become av before indra (SK 89).
Rule. After an eṅ vowel (e or o) at the end of a pada, when a short a follows, the single substitute for both is the pūrvarūpa: the preceding vowel's own form. The word at is a with t after it, and so names only the short a (chapter 1, SK 15).
Example.
hare + ava → the e ends the pada hare and short a follows → the two are replaced by e → hare 'va (not haray ava by SK 61).
viṣṇo + ava → viṣṇo 'va.
Rule. Everywhere — in ordinary speech and in the Veda alike — the word go, ending in eṅ, at the end of a pada, before a short a, optionally keeps its own form unchanged (prakṛti-bhāva, "staying as it is") instead of taking the pūrvarūpa.
Example.
go + agram → o ends the pada go and a follows → left unchanged → go agram; or by SK 86 → go 'gram.
Applies when / blocked by. "Ending in eṅ" excludes citragu + agram → citragvagram: the final of citragu is not o, so it is not eṅ-final. "At the end of a pada" excludes goḥ, a case-form in which the o is inside the word.
Rule. In the opinion of Sphoṭāyana — that is, optionally — the o of go at the end of a pada, before any vowel, is replaced by avaṅ. The substitute is av; the marker ṅ says that it replaces the final sound only. The word at of SK 86 is no longer carried in: any vowel will do.
Example.
go + agram → the pada-final o stands before a vowel → av → gav-agram → by SK 85 a + a → ā → gavāgram.
go + īśaḥ → gav-īśaḥ → by SK 69 → gaveśaḥ (this is why at must not be carried in).
Applies when / blocked by. "At the end of a pada" excludes go + i → gavi: the o is inside the word, and the av there comes from SK 61 instead. The option is a regulated one (vyavasthita-vibhāṣā): usage decides where it is taken, and in gavākṣaḥ ("window", "like the eye of the rays") the av is obligatory.
Rule. Before the word indra, the o of go at the end of a pada is replaced by avaṅ obligatorily: this rule removes the option of SK 88.
Example.
go + indraḥ → av for the o → gav-indraḥ → by SK 69 a + i → e → gavendraḥ.
| SK | Sūtra | What it does | Example |
|---|---|---|---|
| 47 | iko yaṇaci | i u ṛ ḷ → y v r l before a vowel. | sudhī upāsyaḥ → sudhy upāsyaḥ |
| 48 | anaci ca | A yar after a vowel may double when no vowel follows. | sudhy → sudhdhy |
| 49 | sthānivadādeśo'nalvidhau | A substitute is like its original, except in an al-vidhi. | y is not ī for the doubling |
| 50 | acaḥ parasmin pūrvavidhau | …but a vowel-substitute caused by what follows is, for an operation on what precedes. | y would again count as ī |
| 51 | na padāntadvirvacana… | …not in ten operations, doubling among them. | doubling of dh stands |
| 52 | jhalāṃ jaś jhaśi | A jhal before a voiced stop → its jaś. | sudhdhy → suddhy |
| 53 | adarśanaṃ lopaḥ | Non-appearance of a due sound is named lopa. | — |
| 54 | saṃyogāntasya lopaḥ | Final sound of a cluster-final pada drops; vārttikas save the yaṇ and double it. | four forms of suddhy upāsyaḥ |
| 55 | nādinyākrośe putrasya | No doubling in putra before ādinī as an insult. | putrādinī / puttrādinī sarpiṇī |
| 56 | triprabhṛtiṣu śākaṭāyanasya | Three-plus consonants: doubling optional. | inndraḥ / indraḥ |
| 57 | sarvatra śākalyasya | Śākalya: no doubling anywhere. | arkaḥ |
| 58 | dīrghādācāryāṇām | No doubling after a long vowel. | dātram |
| 59 | aco rahābhyāṃ dve | A yar after post-vowel r or h may double. | haryyanubhavaḥ |
| 60 | halo yamāṃ yami lopaḥ | A yam after a consonant may drop before the same yam. | ādityam |
| 61 | eco'yavāyāvaḥ | e o ai au → ay av āy āv before a vowel. | hare + e → haraye |
| 62 | tasya lopaḥ | A marker drops; the y v of ay av are not markers. | haraye, viṣṇave |
| 63 | vānto yi pratyaye | o au → av āv before a y-affix. | gavyam, nāvyam |
| 64 | dhātostannimittasyaiva | …for a root, only if that affix caused the o au. | lavyam; not oyate |
| 65 | kṣayyajayyau śakyārthe | kṣayya, jayya fixed in the sense "capable of". | kṣayyam / kṣeyam |
| 66 | krayyastadarthe | krayya fixed in the sense "for sale". | krayyam / kreyam |
| 67 | lopaḥ śākalyasya | Pada-final y v after a ā may drop before a voiced sound; no further sandhi. | hara ehi / harayehi |
| 68 | ekaḥ pūrvaparayoḥ | Heading: one substitute for both vowels. | — |
| 69 | ādguṇaḥ | a ā + vowel → one guṇa. | upa + indraḥ → upendraḥ |
| 70 | uraṇ raparaḥ | An aṇ replacing ṛ comes with r (ḷ: with l). | kṛṣṇarddhiḥ, tavalkāraḥ |
| 71 | jharo jhari savarṇe | A jhar after a consonant may drop before a savarṇa jhar. | kṛṣṇardhiḥ / kṛṣṇarddhiḥ / kṛṣṇardddhiḥ |
| 72 | vṛddhireci | a ā + e o ai au → one vṛddhi. | kṛṣṇa + ekatvam → kṛṣṇaikatvam |
| 73 | etyedhatyūṭhsu | Vṛddhi before eti, edhati, ūṭh; five vārttikas. | upaiti, praṣṭhauhaḥ, akṣauhiṇī |
| 74 | upasargādṛti dhātau | Prefix + ṛ-initial root → vṛddhi ār. | upa + ṛcchati → upārcchati |
| 75 | antādivacca | The ekādeśa counts as final of the first and initial of the second. | r of upārcchati is pada-final |
| 76 | kharavasānayorvisarjanīyaḥ | Pada-final r → visarga before khar or pause; not for an r final by SK 75. | upārcchati keeps its r |
| 77 | vā supyāpiśaleḥ | Prefix + ṛ-initial noun-derived root → vṛddhi optional. | prārṣabhīyati / prarṣabhīyati |
| 78 | eṅi pararūpam | Prefix + e o-initial root → the second vowel alone. | pra + ejate → prejate |
| 79 | aco'ntyādi ṭi | Last vowel plus what follows = ṭi; pararūpa of the ṭi in śakandhu etc. | śakandhuḥ, sīmantaḥ, manīṣā |
| 80 | omāṅośca | a ā + om or āṅ → the second alone. | śivāyoṃ namaḥ, śivehi |
| 81 | avyaktānukaraṇasyāta itau | The a of a noise-word + the i of iti → i alone. | paṭat + iti → paṭiti |
| 82 | nāmreḍitasyāntyasya tu vā | Not for an āmreḍita; its final t optionally. | paṭatpaṭeti |
| 83 | tasya paramāmreḍitam | Second member of a doubled word = āmreḍita. | paṭat paṭat |
| 84 | jhalāṃ jaśo'nte | Pada-final jhal → jaś. | paṭatpaṭad iti, śrad iti |
| 85 | akaḥ savarṇe dīrghaḥ | a i u ṛ ḷ + same-kind vowel → one long vowel. | daitya + ariḥ → daityāriḥ |
| 86 | eṅaḥ padāntādati | Pada-final e o + short a → the first alone. | hare + ava → hare 'va |
| 87 | sarvatra vibhāṣā goḥ | Pada-final go before a may stay unchanged. | go agram / go 'gram |
| 88 | avaṅ sphoṭāyanasya | Pada-final o of go before a vowel → av, optionally. | gavāgram, gaveśaḥ |
| 89 | indre ca | …before indra, obligatorily. | go + indraḥ → gavendraḥ |