This chapter opens the second half (uttarārdha) of the Kaumudī. In the verses that open this half, Bhaṭṭoji says the first half taught the affixes of books 4 and 5 of the Aṣṭādhyāyī; the affixes of book 3 are now told. The verbs come first. A finished verb, such as bhavati "he is", is a tiṅanta: a word ending in a verb-ending (tiṅ). Before the first rule, the printed Kaumudī lists the ten lakāras, the ten abstract affixes that carry tense and mood: laṭ, liṭ, luṭ, lṛṭ, leṭ, loṭ, laṅ, liṅ, luṅ, lṛṅ. Of these the fifth, leṭ, is found only in the Veda. After this chapter you will be able to take one root, bhū "to be", through every lakāra, and derive bhavati, babhūva, bhavitā, bhaviṣyati, bhavatu, abhavat, bhavet, bhūyāt, abhūt and abhaviṣyat. Bhaṭṭoji introduces each rule at the point in the derivation where it first bites, so the rules come from all over the Aṣṭādhyāyī. You will then take the root edh through the same lakāras with the other set of endings: edhate, edhāṃcakre, edhatām, aidhata, edheta, edhiṣīṣṭa. Finally you will walk the list of roots that begin with bhū, the bhvādi class, about a thousand roots, and learn the special rules each group needs: gopāyati, kāmayate, acīkamata, sīdati, paśyati, iyāja and more. The chapter ends with ṛtīyate. It has 272 sūtras, SK 2151 to SK 2422.
bhū + laṭ → bhū + l → bhū + tipThis group sets up the machinery. A lakāra is chosen by the time or mood of the action. SK 2151 gives laṭ for the present. SK 2152 says what a lakāra denotes: the agent, the object, or the bare action. SK 2153 makes the l of the lakāra the thing replaced. SK 2154 lists the eighteen verb-endings that replace it. The running example is bhū "to be", in the present: bhū + laṭ, then bhū + l, then one of the eighteen, bhū + tip.
Rule. The affix laṭ comes after a root whose action is present (vartamāna). The a and the ṭ of laṭ are it, so only l remains.
The heading dhātoḥ "after a root" is in force here. It is SK 2829, taught much later in the Kaumudī. It governs every rule to the end of the third book of the Aṣṭādhyāyī, so every lakāra rule in this chapter comes "after a root".
Example.
bhū "being", said of the present
→ bhū + laṭ (this rule)
→ the a and ṭ are it and are dropped → bhū + l.
Applies when / blocked by. "Present" describes the action of the root, not the agent. Bhaṭṭoji's commentators both say so. Otherwise pacati "he cooks" could be said of a cook who is present, though his cooking is past. A note in the printed edition defines presentness: being the substrate of an action begun and not yet finished.
Bhaṭṭoji's commentators both say that laṭ only makes the present manifest; it is a dyotaka, a signifier. One of them adds that the other view, that laṭ denotes the present, is also tenable. What laṭ does denote is taught in the next rule.
Watch out. Why is the a of laṭ an it, and not merely there for pronunciation? Bhaṭṭoji's commentators both answer. The a tells laṭ apart from liṭ, luṭ and the rest. Since it must be removed, it must be an it.
Rule. This rule says what a lakāra denotes. It is two statements. After a root that takes an object (sakarmaka), the lakāras denote the object (karman), and also the agent (kartṛ). After a root that takes no object (akarmaka), they denote bhāva, the bare action itself, and also the agent.
The word kartari "in the sense of the agent" is drawn in by ca "and". It comes from SK 2832, kartari kṛt, taught later. An akarmaka root never has a lakāra in the sense of the object, since it has none.
Example.
bhū "to be" takes no object; it is akarmaka.
→ its l may denote the agent, or the bare action.
→ every form of bhū in this part of the chapter is in the sense of the agent: bhavati "he is".
Applies when / blocked by. Why is the first statement limited to sakarmaka roots? One of Bhaṭṭoji's commentators explains. Otherwise a sakarmaka root could take a lakāra of bare action. Then in devadattena ghaṭaṃ kriyate the object "pot" would not be expressed and would take the second case, which is wrong.
A root whose object is not meant to be expressed also counts as akarmaka here. One of Bhaṭṭoji's commentators gives devadattena bhujyate "there is eating by Devadatta": the rice he eats is there, but it is not meant to be expressed. The other, citing Haradatta, gives neha bhujyate "there is no eating here". He calls this the stronger view; scholars discuss this.
Rule. This is an adhikāra, a heading. It means "in place of l". The rules after it teach substitutes for the l of a lakāra.
The a of la is not meant. So all ten lakāras are covered: laṭ, liṭ, luṭ and the rest, which all reduce to l.
Example.
bhū + l → the l is the sthānin, the thing to be replaced
→ every verb-ending in this chapter stands in its place.
Applies when / blocked by. Why is there no substitute for the l of cūḍāla or lunāti? One of Bhaṭṭoji's commentators answers that only an l introduced after a root is meant, since dhātoḥ is the heading. The other gives two answers. The first: only an l taught in the senses of SK 2152 is meant. The second is the same as the first commentator's.
Rule. Eighteen substitutes (ādeśa) come in place of the l. They are these, in threes:
Together they are the tiṅ, the verb-endings. The name is a pratyāhāra: from the ti of tip to the ṅ of mahiṅ.
Example.
bhū + l → one of the eighteen comes in place of l
→ which one is decided by SK 2155–2165
→ for "he is", bhū + tip, and tip is heard as ti.
Applies when / blocked by. One of Bhaṭṭoji's commentators notes three points. The iṭ in this list is a substitute, not the augment iṭ. Its ṭ serves a later rule, SK 2257. The ṅ of mahiṅ marks the whole set, not mahi alone. It gives the pratyāhāras tiṅ and taṅ.
The other commentator explains the p of tip, sip and mip. It keeps them from being treated as ṅit, that is, as marked with ṅ, under SK 2234 (sārvadhātukam apit). That matters in forms like dveṣṭi.
Watch out. mib in the sūtra is only the sandhi form of mip. The list has eighteen items, not seventeen.
bhū + l → bhū + (tip … mas)The eighteen endings fall into two sets of nine. The first nine, tip to mas, are parasmaipada. The second nine, ta to mahiṅ, are ātmanepada. The root decides which set it takes. SK 2155–2156 give the two names. SK 2157–2158 say which roots take ātmanepada. SK 2159 gives parasmaipada to all the rest. bhū is one of the rest, so it takes the first nine: bhū + (tip … mas).
Rule. The substitutes of l (lādeśa) are named parasmaipada. As first given, the name covers all eighteen. SK 2156 then takes nine of them away.
Example.
tip, tas, jhi … ta, ātām … mahiṅ: all eighteen are named parasmaipada by this rule
→ SK 2156 names nine of them ātmanepada
→ parasmaipada is left with tip … mas.
Applies when / blocked by. A note in the printed edition says what "substitutes of l" means. It means substitutes taught for l as such. So an affix like sic, taught in place of something else, is not parasmaipada.
Rule. taṅ and āna are named ātmanepada. taṅ is the pratyāhāra of the nine endings from ta to mahiṅ. āna means the two affixes śānac and kānac, which are also substitutes of l.
This is an apavāda, an exception, to SK 2155. Both names are taught in a section of the Aṣṭādhyāyī where a thing may bear only one name. A note in the printed edition gives this reason. So the nine taṅ lose the name parasmaipada. Parasmaipada is left with the nine from tip to mas.
Example.
ta, ātām, jha, thās, āthām, dhvam, iṭ, vahi, mahiṅ → ātmanepada
tip, tas, jhi, sip, thas, tha, mip, vas, mas → parasmaipada.
Applies when / blocked by. Only the āna that replaces l is meant. Bhaṭṭoji's commentators both say so. There is another affix cānaś (SK 3109) which does not replace l. So it is not ātmanepada, and it can come even after a root that takes parasmaipada: nighnānāḥ.
Rule. Ātmanepada comes in place of l after two kinds of root. The first is a root whose it is an anudātta vowel (anudāttet). The second is a root that ends in something ṅit, marked with ṅ, in its original statement (upadeśa).
Two limits apply to the second kind only. "In upadeśa" means the ṅ must be there in the original statement. "Ends in" means a root ending in a ṅit affix counts too. This reading, "that which ends in it", is tadantavidhi.
Example.
bhū has neither an anudātta it nor a ṅ
→ this rule does not apply to it.
The ātmanepada roots, beginning with edh, come in the next part of the chapter.
Applies when / blocked by. Bhaṭṭoji's commentators give counter-examples. Why "in upadeśa"? In cukuṭiṣati the affix san is ṅit only by extension from another rule, not in its original statement. So no ātmanepada. Why "ends in"? So that bobhūyate, a root ending in the ṅit affix yaṅ, takes ātmanepada.
How is "after a root" obtained here? Bhaṭṭoji's commentators differ, and scholars discuss this. Both agree that the sense is "after a root".
Rule. After a root whose it is a svarita vowel (svaritet), or whose it is ñ (ñit), ātmanepada comes. The condition is that the fruit of the action (kriyāphala) goes to the agent.
Example.
bhū has neither a svarita it nor ñ
→ this rule does not apply to it.
Applies when / blocked by. What is "the fruit"? One of Bhaṭṭoji's commentators says it is what the action is undertaken for. In hotā yājyayā yajati "the hotṛ priest sacrifices with the yājyā verse", the fruit is heaven. It goes to the patron of the sacrifice, not to the priest. So the priest's verb takes no ātmanepada. A fee is not the fruit; the printed edition's note agrees.
When the fruit does not go to the agent, this rule gives nothing. Such a root then falls under SK 2159.
Rule. After any other root (śeṣa, "the rest"), parasmaipada comes in the sense of the agent. "The rest" means a root with no condition for ātmanepada: nothing that SK 2157 or SK 2158 asks for.
Example.
bhū: no anudātta it, no ṅ, no svarita it, no ñ
→ no condition for ātmanepada
→ parasmaipada in the sense of the agent → bhū + (tip … mas).
Applies when / blocked by. One of Bhaṭṭoji's commentators says this is the rule that gives bhū parasmaipada. It still leaves open which of the nine comes. All nine might come at once, or in turn. So Bhaṭṭoji first gives the names of the three persons (SK 2160), then the rules that choose between them.
bhū + (tip … mas) → bhū + tibhū now has nine endings to choose from. This group narrows them to one. SK 2160 names the three triads prathama, madhyama and uttama; together they are the three persons (puruṣa). SK 2161 names the three members of each triad singular, dual and plural. SK 2162–2165 then choose the person: madhyama with "you", uttama with "I", prathama elsewhere. For "he is", no "you" or "I" is meant and the agent is one. So the choice is tip, heard as ti: bhū + ti.
Rule. In each of the two sets, parasmaipada and ātmanepada, the tiṅ fall into three triads. In order, they are named prathama, madhyama and uttama. The earlier teachers call these names puruṣa, "person".
The words "parasmaipada" and "ātmanepada" are carried forward from SK 2155 and SK 2156. That is why Bhaṭṭoji writes "in both padas", that is, in parasmaipada and in ātmanepada.
Example.
parasmaipada: tip tas jhi = prathama · sip thas tha = madhyama · mip vas mas = uttama
ātmanepada: ta ātām jha = prathama · thās āthām dhvam = madhyama · iṭ vahi mahiṅ = uttama.
Applies when / blocked by. Why "in both padas"? Bhaṭṭoji's commentators both answer. There are six triads but only three names. Without the two padas, the names could not be paired off one by one (yathāsaṃkhya, SK 128). Each triad would get all three names.
Watch out. The names follow the order of the list, not "first, second, third person" in the English sense. prathama "first" is the triad used where there is no "you" or "I" (SK 2165). uttama "last" is the triad used for "I" (SK 2164).
Rule. Within each triad that has received the name prathama, madhyama or uttama, the three members are named one by one (ekaśaḥ). They are ekavacana (singular), dvivacana (dual) and bahuvacana (plural).
Example.
tip = prathama ekavacana · tas = prathama dvivacana · jhi = prathama bahuvacana
sip = madhyama ekavacana … mas = uttama bahuvacana.
Applies when / blocked by. These names let the number rules of chapter 8 choose within a triad. One of Bhaṭṭoji's commentators says so. SK 186 gives the dual and singular for two and one; SK 187 gives the plural for many. Without the names, the three endings would come all at once or in turn.
The other commentator explains Bhaṭṭoji's phrase "which have received the names prathama and the rest". The two sets of names must hold together on the same ending. Otherwise, under the rule "one name only", they would exclude each other. Then in the imperative of ad "eat", the ending ni (SK 2203) could lose the name uttama when the singular is meant. The augment of SK 2204 would not come, and a wrong form adni would pass.
Rule. The madhyama comes in place of l when yuṣmad "you" stands nearby. yuṣmad must have the same referent (samānādhikaraṇa) as the l: it must denote the very kāraka that the verb-ending denotes. It may be actually used, or only understood (sthānin).
Here upapada is the plain word "a word uttered nearby", not the technical term. A note in the printed edition says so.
Example.
tvaṃ bhavasi "you are": tvam "you" is the agent, and the verb-ending also denotes the agent
→ same referent → madhyama → singular: bhū + sip
→ by SK 2167–2168 → bhavasi.
With yuṣmad only understood: rāma pāhi "Rāma, protect!" is madhyama, with no tvam spoken. This example is from one of Bhaṭṭoji's commentators.
Applies when / blocked by. "Same referent" is essential. One of Bhaṭṭoji's commentators gives tvāṃ paśyati "he sees you", tvayā kriyate "it is done by you" and tubhyaṃ dadāti "he gives to you". There yuṣmad is the object, the instrument or the recipient, not the agent the ending denotes. So the verb stays prathama.
The other commentator adds that the respectful bhavat "your honour" is not yuṣmad. So bhavān āgacchati "your honour comes" is prathama.
Rule. This rule works in mockery (prahāsa), and it is two statements. First: a root that has the verb man "think" nearby takes the madhyama, where "I" is meant and the uttama was due. Second: man itself takes the uttama, where "you" is meant and the madhyama was due. The uttama of man is always singular, even for two or many.
The man meant is manyate "he thinks", with the affix śyan between root and ending (a vikaraṇa, SK 2167). The rule swaps the two persons, madhyama and uttama.
Example.
Bhaṭṭoji gives the examples at SK 2170, once the present forms exist. The plain sentence would be:
ehi, manyase "odanaṃ bhokṣye" "Come, you think 'I shall eat rice'".
In jest the persons are swapped:
bhokṣye "I shall eat" → madhyama → bhokṣyase
manyase "you think" → uttama singular → manye
→ ehi manye odanaṃ bhokṣyase iti, bhuktaḥ so 'tithibhiḥ "Come! You think, 'I shall eat rice' — it has been eaten by the guests."
One of Bhaṭṭoji's commentators says it is teasing a son-in-law who comes to eat after everyone has eaten.
Applies when / blocked by. The rule needs "you" to be nearby, as in SK 2162. So it does not apply with bhavat (SK 2170). It also needs mockery. A plain statement keeps the normal persons (SK 2170).
Rule. The uttama comes when asmad "I, we" stands nearby with the same referent. As in SK 2162, asmad must denote the kāraka the ending denotes. It may be used or only understood.
Example.
ahaṃ bhavāmi "I am": aham "I" is the agent, and the ending denotes the agent
→ same referent → uttama → singular: bhū + mip
→ by SK 2167–2170 → bhavāmi.
Applies when / blocked by. Bhaṭṭoji writes only "when asmad is such". His commentators both fill it in from SK 2162: denoting the kāraka the ending denotes, used or not used.
Rule. In the remaining case (śeṣa), the prathama comes. The remaining case is outside the scope of SK 2162–2164: no "you" and no "I" with the same referent.
Bhaṭṭoji then sets up the model root: bhū (root 1 of the list), in the sense sattā "being". One of Bhaṭṭoji's commentators says sattā here means remaining in one's own form.
Example.
Bhaṭṭoji says only "when the agent is meant and bhū + ti stands". One of his commentators spells out the choice for "he is":
the agent is meant, and the action is present → laṭ (SK 2151, 2152)
→ no condition for ātmanepada → parasmaipada (SK 2159)
→ no "you" or "I" with the same referent → prathama (this rule)
→ one agent → singular: tip (SK 2161, SK 186)
→ bhū + ti.
Applies when / blocked by. One of Bhaṭṭoji's commentators gives two mixed cases. In tvam ahaṃ ca pacāvaḥ "you and I cook", both "you" and "I" are present. The uttama wins, because SK 2164 is the later rule. In devadattas tvaṃ ca pacatha "Devadatta and you cook", "you" is present. So the case is not a "remainder", and the madhyama comes, not the prathama.
bhū + ti → bhū + a + ti → bhavatiNow the ending is chosen, the root is built up to meet it. SK 2166 names the verb-endings sārvadhātuka. SK 2167 puts the affix śap between root and ending; only its a is heard. SK 2168 turns the ū of bhū into its guṇa o, and o becomes av before the vowel. So bhū + ti becomes bhū + a + ti, then bhavati "he is". SK 2169 gives ant for the jh of jhi (bhavanti). SK 2170 lengthens the a before mi, vas, mas (bhavāmi). The group ends with the whole present of bhū.
Rule. The tiṅ endings, and affixes marked with ś (śit), are named sārvadhātuka, if they are taught under the heading dhātoḥ "after a root".
Example.
bhū + ti: ti is a tiṅ taught after a root → sārvadhātuka.
śap (SK 2167) is marked with ś → also sārvadhātuka.
Applies when / blocked by. Why "taught after a root"? Bhaṭṭoji's commentators both answer with harīn "the Haris" (second case plural). Its ending śas is also marked with ś. If it were sārvadhātuka, it would count as ṅit by SK 2234. Then SK 245 (gher ṅiti) would give guṇa, and the form would be wrong. śas is a case-ending, not taught after a root, so it is not sārvadhātuka.
Rule. The affix śap comes after a root when a sārvadhātuka follows that denotes the agent. Its ś and p are it (SK 195, SK 1), so only a remains.
An affix placed between a root and its ending, like śap, is called a vikaraṇa. The name is from the earlier teachers.
Example.
bhū + ti → ti is sārvadhātuka and denotes the agent
→ bhū + śap + ti → bhū + a + ti.
Applies when / blocked by. "Denoting the agent" matters. A note in the printed edition says that in the sense of the object or the bare action, the affix yak (SK 2756) blocks śap anyway. So the word "agent" is there mainly for the rules that follow.
Rule. When a sārvadhātuka or an ārdhadhātuka follows, guṇa replaces the final ik of an aṅga ending in an ik. (The name ārdhadhātuka is taught at SK 2187.)
The word guṇa is carried from SK 2346. The ik is supplied by SK 34.
Example.
bhū + a + ti: the aṅga bhū ends in the ik ū; a (śap) is sārvadhātuka
→ guṇa of ū is o → bho + a + ti
→ o becomes av before a vowel (SK 61) → bhav + a + ti
→ bhavati "he is".
With tas: bhū + a + tas → bhavatas → s becomes ru, then visarga → bhavataḥ "the two are".
Rule. The jh that is part of an affix is replaced by ant. The final a of anta in the sūtra is only for pronunciation.
The word "affix" is carried from SK 475. That rule also says "initial", but "initial" is not carried. So jh anywhere in an affix is replaced.
Example.
bhū + jhi → śap, guṇa, av → bhava + jhi
→ jh becomes ant (this rule) → bhava + anti
→ a + a gives one a, the later one (pararūpa, SK 191) → bhavanti "they are".
The madhyama forms are made like bhavati: bhavasi (sip), bhavathaḥ (thas), bhavatha (tha).
Applies when / blocked by. "Part of an affix" matters. In ujjhitā the jh belongs to the root, so it stays. This is from one of Bhaṭṭoji's commentators.
Watch out. Why is the a of bhava not lengthened before anti by SK 2170? The jh of jhi is a yañ sound, and SK 2170 lengthens before a yañ. But ant is not itself a yañ. And SK 49, the rule that a substitute is treated like what it replaced, does not apply when the rule depends on a particular sound. So ant is not treated as the yañ jh. One of Bhaṭṭoji's commentators gives this reason. So SK 191 applies: bhavanti, not bhavānti.
Rule. The a at the end of an aṅga is made long when a sārvadhātuka beginning with a yañ sound follows. yañ is the pratyāhāra y v r l ñ m ṅ ṇ n jh bh. The word "sārvadhātuka" is carried from SK 2444.
Example.
bhū + mi → śap, guṇa, av → bhava + mi
→ m is a yañ → a made long → bhavāmi "I am".
bhava + vas → bhavāvaḥ "we two are"; bhava + mas → bhavāmaḥ "we are".
But is bhava the aṅga? The affix mi was added after the root bhū, not after bhava. One of Bhaṭṭoji's commentators answers: by the word "beginning with" in SK 199, what includes the vikaraṇa is also an aṅga.
Bhaṭṭoji then gives the whole present with its pronouns:
sa bhavati "he is" · tau bhavataḥ "those two are" · te bhavanti "they are"
tvaṃ bhavasi "you are" · yuvāṃ bhavathaḥ "you two are" · yūyaṃ bhavatha "you all are"
ahaṃ bhavāmi "I am" · āvāṃ bhavāvaḥ "we two are" · vayaṃ bhavāmaḥ "we are".
Applies when / blocked by. Here Bhaṭṭoji also gives the mockery examples of SK 2163. The singular example is ehi manye odanaṃ bhokṣyase iti, bhuktaḥ so 'tithibhiḥ. For two or many, the verb "eat" changes but manye stays singular:
etaṃ manye odanaṃ bhokṣyethe "come, you two — you think 'we two shall eat rice'"
eta manye odanaṃ bhokṣyadhve "come, all of you — you think 'we shall eat rice'".
Bhaṭṭoji prints the two together as etam eta vā …; one of his commentators reads them as these two sentences.
The real sense is bhokṣye, bhokṣyāvahe, bhokṣyāmahe "I, we two, we shall eat", and manyase, manyethe, manyadhve "you, you two, you all think".
Two counter-examples follow. etu bhavān, manyate "odanaṃ bhokṣye" iti has bhavān "your honour", not yuṣmad. The condition of SK 2162 is carried into SK 2163, so no swap. ehi, manyase "odanaṃ bhokṣye" iti is a plain statement, with no mockery. So the persons stay as they are.
bhū + liṭ → bhū + ṇal → bhū + a → bhūv + aThe second lakāra is liṭ, for a past beyond the speaker's sight. This group gets bhū as far as bhūv + a. SK 2171 gives liṭ. SK 2172 makes its endings ārdhadhātuka only, so śap does not come and the root stays bare. SK 2173 gives liṭ its own parasmaipada endings: tip becomes ṇal, heard as a. SK 2174 adds v to bhū before that vowel. The result is bhūv + a. Group I.F doubles it into babhūva "he was".
Rule. liṭ comes after a root whose meaning is past (bhūta), not of today (anadyatana), and beyond perception (parokṣa). Then, as always, tip and the rest replace the l.
"Past" is the heading bhūte (SK 2995, taught later). "Not of today" is carried from SK 2205.
Example.
bhū, of a being long past and not witnessed
→ bhū + liṭ (this rule) → bhū + l → prathama singular → bhū + tip.
Applies when / blocked by. What is "today"? One of Bhaṭṭoji's commentators gives the definition he finds in Kaiyaṭa's commentary. "Today" is the day together with the last watch of the past night and the first watch of the coming night.
What is "beyond perception"? The same commentator says all views agree on this much: it is outside the range of the speaker's senses. The other commentator says every action is imperceptible, so what is meant is that the means of the action were not perceived; scholars discuss this.
Rule. A tiṅ that replaces liṭ is named ārdhadhātuka only (eva), not sārvadhātuka. So śap and the other vikaraṇas, which need a sārvadhātuka, do not come.
The word "only" is carried from SK 2463. Without it the ending would have both names: sārvadhātuka as a tiṅ (SK 2166), and ārdhadhātuka by this rule. One of Bhaṭṭoji's commentators explains that the two names could otherwise stand together. The name sārvadhātuka is taught outside the one-name section of SK 2156, so it does not exclude another name.
Example.
bhū + tip (liṭ) → tip is ārdhadhātuka only
→ no śap → the base stays the bare root bhū, not bhava.
Rule. The nine parasmaipada endings of liṭ are replaced, one by one, by nine others:
The word "of liṭ" is carried from SK 2241.
Example.
bhū + tip → tip is replaced by ṇal
→ ṇ and l are it → bhū + a.
Applies when / blocked by. ṇal replaces the whole of tip. One of Bhaṭṭoji's commentators explains why. Before it takes its place, ṇal is not yet an affix. So its ṇ is not yet an it by SK 189, and ṇal counts as more than one sound (SK 45). The ṇ is there for vṛddhi in forms like jagāda. The l is for the accent.
Watch out. The a that replaces tha also replaces the whole. Bhaṭṭoji's commentators both report the explanation: the sūtra writes two a's merged into one. So it is a substitute of more than one sound.
Rule. bhū takes the augment vuk when a vowel of luṅ or liṭ follows. An augment (āgama) is a sound added to a form, not in place of anything. In vuk, the k is it and the u is for pronunciation, so v is added. Being marked with k, it goes at the end of bhū (SK 36).
"When a vowel follows" is carried from SK 271.
Example.
bhū + a (liṭ, from ṇal) → a vowel of liṭ follows
→ vuk → bhūv + a.
Applies when / blocked by. vuk is nitya, "constant". It would apply whether or not guṇa or vṛddhi had applied first. So it blocks guṇa by SK 2168 and vṛddhi by SK 254, which would otherwise come here. Bhaṭṭoji states this; one of his commentators explains it.
Why "when a vowel follows"? In the luṅ form abhūt "he was", a consonant follows. So there is no vuk. Bhaṭṭoji's commentators both give this.
bhūv a → bhūv bhūv a → bhu bhūv a → babhūvaliṭ doubles the root. This group teaches how, and then finishes the liṭ of bhū. SK 2175–2176 are headings for doubling. SK 2177 doubles the root before liṭ: bhūv bhūv a. SK 2178 names the first copy abhyāsa. SK 2179–2182 trim the abhyāsa: bhū, then bhu, then bha, then ba. So babhūva. SK 2183 explains why the second ū does not change. SK 2184 adds i before some endings: babhūvitha, babhūvima.
Rule. This is a heading. It says: "two utterances of the first single-vowelled part (ekāc)". An ekāc is a part with one vowel. The rules that follow say when this doubling happens.
The heading runs up to, but not including, SK 1003 (chapter 25). Bhaṭṭoji gives this sūtra no gloss of its own. He treats it together with SK 2176.
Example.
bhūv a → the first single-vowelled part is bhūv
→ doubled when a rule under this heading applies (SK 2177) → bhūv bhūv a.
Applies when / blocked by. Why only a heading? One of Bhaṭṭoji's commentators says it would apply far too widely as a rule by itself.
Watch out. This doubling is not substitution. One of Bhaṭṭoji's commentators, following the Bhāṣya, says "two" means two utterances: the part is simply said twice (dviḥprayoga). Contrast chapter 43, where sarvasya dve puts two word-forms in place of one.
Rule. This is a heading too, read with SK 2175. In a root that begins with a vowel, the second single-vowelled part is doubled: the one after the initial vowel. The word "single-vowelled part" is supplied.
Example.
bhū begins with a consonant, so this rule does no work on it.
Its use comes with the vowel-initial roots later in the chapter.
Applies when / blocked by. ajādeḥ means "after the initial vowel". Bhaṭṭoji's commentators both insist on this reading, and give a reason. If it meant "of what begins with a vowel", then SK 2446 would wrongly stop the doubling of the d in indidrīyiṣati.
Rule. When liṭ follows, the first single-vowelled part of the root is uttered twice. That part must not itself be an abhyāsa (SK 2178). In a root that begins with a vowel, it is the second single-vowelled part, the one after the initial vowel.
Example.
bhūv a has four single-vowelled parts: bhū, ūv, ū, and the whole bhūv
→ the whole is doubled, and its parts go with it
→ bhūv bhūv a.
One of Bhaṭṭoji's commentators gives the image: when a tree moves, all its parts move.
Applies when / blocked by. Why "of the root"? Otherwise the first single-vowelled part of anything before liṭ would double, and jajāgāra would fail. Why "not an abhyāsa"? It stops a second doubling when san follows a yaṅ-form, or caṅ follows a san-form (SK 2395, SK 2315). Both points are from one of Bhaṭṭoji's commentators.
Rule. Of the two copies made in this doubling section, the former is named abhyāsa.
Example.
bhūv bhūv a → the first bhūv is the abhyāsa
→ [bhūv] bhūv a, where the brackets mark the abhyāsa.
Applies when / blocked by. "In this section" means under SK 2175. Bhaṭṭoji's commentators both say so. One of them adds that doubling by SK 2139 (sarvasya dve, chapter 43) therefore makes no abhyāsa.
Rule. In the abhyāsa, only the initial consonant (hal) remains. The other consonants are deleted. "Of the abhyāsa" is carried from SK 2620.
Example.
[bhūv] bhūv a → the abhyāsa has two consonants, bh and v
→ bh is initial and remains; v is deleted
→ [bhū] bhūv a.
Rule. The vowel of the abhyāsa is made short (hrasva). The word "vowel" comes in by SK 35.
Example.
[bhū] bhūv a → ū made short
→ [bhu] bhūv a.
Applies when / blocked by. For bhū, SK 2181 will replace this vowel anyway. One of Bhaṭṭoji's commentators says the rule is still needed for forms like lulāva. It is taught here because this is its place in the derivation.
Rule. When liṭ follows, a replaces the u of the abhyāsa of bhū. bhavateḥ is the way the root bhū is cited. By SK 42, the a replaces the final vowel of the abhyāsa. "When liṭ follows" is carried from SK 2353.
Example.
[bhu] bhūv a → the u of the abhyāsa becomes a
→ [bha] bhūv a.
Applies when / blocked by. The rule holds for the object and bare-action senses too, not only the agent. One of Bhaṭṭoji's commentators cites the poets' bibharāṃbabhūve for this.
Rule. In the abhyāsa, a jhal consonant is replaced by a car, and also (ca) by a jaś. Each takes its nearest match (SK 39):
The substitute is always of the same class as the sound it replaces. The word jhal is carried from SK 52.
Example.
[bha] bhūv a → bh is a jhaś
→ its jaś of the same class is b
→ [ba] bhūv a → babhūva "he was".
Applies when / blocked by. A sound that is already a jaś is "replaced" by itself. One of Bhaṭṭoji's commentators says the rule falls like rain, even where it changes nothing. Bhaṭṭoji puts it in general terms: a jaś by nature gets a jaś by nature, a car by nature gets a car.
A note in the printed edition adds that changes to the abhyāsa do not block one another. So SK 2179 and SK 2181 both apply to the same abhyāsa.
Rule. This is a heading. It opens a zone of rules: from the next rule of the Aṣṭādhyāyī, SK 2544, to the end of the pāda, including the bha section (SK 233). A pāda is a quarter of one of the eight books (adhyāya) of the Aṣṭādhyāyī. Bhaṭṭoji calls these operations ābhīya, "of the zone up to bha". Within the zone, an ābhīya operation is treated as not having taken effect (asiddha). That holds when another ābhīya operation with the same basis is to be made.
Example.
The problem: vuk (SK 2174) is ābhīya, and so is uvaṅ (SK 271).
babhūv + a (after doubling, SK 2177–2182) → vuk is asiddha for uvaṅ → uvaṅ sees babhū + a, with a vowel after ū
→ uvaṅ would turn ū into uv → babhuv a
→ guṇa of the light penultimate u (SK 2189, below) → babhov a
→ the wrong form babhova.
The fix is a vārttika, a supplementary statement: vugyuṭāv uvaṅyaṇoḥ siddhau vaktavyau. "vuk and yuṭ are to be treated as having taken effect, for uvaṅ and for yaṇ, respectively." The second pair is not needed for bhū: yuṭ is another augment, and yaṇ is the class of semivowels y v r l.
babhūv + a → vuk counts as done → no vowel follows ū → no uvaṅ
→ babhūva "he was"
tas → atus → babhūvatuḥ "the two were"
jhi → us → babhūvuḥ "they were".
Applies when / blocked by. vuk still has its other work. It blocks guṇa and vṛddhi, as SK 2174 said. One of Bhaṭṭoji's commentators notes this.
Watch out. What exactly is "asiddha" here, the operation or the rule that teaches it? Bhaṭṭoji's commentators differ, and scholars discuss this. For babhūva the result is the same.
Rule. An ārdhadhātuka that begins with a val consonant takes the augment iṭ. val is every consonant except y. iṭ is heard as i; being marked with ṭ, it goes at the front of the affix (SK 36).
Example.
madhyama:
sip → thal → bhū + tha → th is a val → iṭ → bhū + i + tha → vuk, doubling → babhūvitha "you were"
thas → athus (begins with a vowel, no iṭ) → babhūvathuḥ "you two were"
tha → a → babhūva "you all were"
uttama:
mip → ṇal → babhūva "I was"
vas → va → v is a val → iṭ → babhūviva "we two were"
mas → ma → iṭ → babhūvima "we were".
Applies when / blocked by. Does SK 2381 (śryukaḥ kiti) forbid iṭ in babhūviva? Bhaṭṭoji's commentators both say no. SK 2293, a restriction on a few roots, lets every other root take iṭ in liṭ.
Why "ārdhadhātuka"? In jugupsati the affix san is not taught under the heading "after a root". So it is not ārdhadhātuka, and takes no iṭ. This is from one of Bhaṭṭoji's commentators.
Watch out. babhūva is three forms: prathama singular, madhyama plural and uttama singular. They come from ṇal, a and ṇal.
bhū + tās + ti → bhū + i + tās + ḍā → bhavit ā → bhavitāSanskrit has two futures. luṭ is for a future not of today; lṛṭ is the general future. SK 2185 gives luṭ. SK 2186 gives each future its own vikaraṇa: tās for luṭ, sya for lṛṭ. SK 2187 names these affixes ārdhadhātuka, so they take iṭ. SK 2188–2192 build the forms of luṭ. Its prathama endings become ḍā, rau, ras, and parts of tās drop away. So bhū + tās + ti becomes bhavitā "he will be". SK 2193 gives lṛṭ: bhaviṣyati "he will be".
Rule. luṭ comes after a root in the sense of a future (bhaviṣyat) that is not of today (anadyatana). "Future" is carried from SK 3171, taught later. The u and ṭ of luṭ are it.
Example.
bhū, of a being that will come about after today
→ bhū + luṭ → bhū + l → bhū + tip, heard as bhū + ti.
Applies when / blocked by. "Not of today" was explained under SK 2171. One of Bhaṭṭoji's commentators refers to SK 1299 for the definition of "today".
Rule. After a root, sya comes before lṛ, and tāsi before luṭ. lṛ covers both lṛṭ and lṛṅ. The pairing is one by one (yathāsaṃkhya): sya with lṛ, tāsi with luṭ. The i of tāsi is for pronunciation, so the affix is tās.
This is an exception (apavāda) to śap (SK 2167) and the other vikaraṇas.
Example.
luṭ: bhū + ti → tās instead of śap → bhū + tās + ti.
lṛṭ: bhū + ti → sya instead of śap → bhū + sya + ti (SK 2193).
Applies when / blocked by. lṛ names two lakāras, so there are three lakāras for two affixes. A note in the printed edition says the pairing still holds, since the sūtra names two items on each side.
Rule. Any other affix taught under the heading "after a root" is named ārdhadhātuka. "Other" means other than the tiṅ and śit affixes, which SK 2166 names sārvadhātuka.
Example.
bhū + tās + ti → tās is taught after a root, and is not tiṅ or śit
→ tās is ārdhadhātuka
→ it begins with t, a val → iṭ by SK 2184 → bhū + i + tās + ti.
Applies when / blocked by. Bhaṭṭoji's commentators both give jugupsate. Its san is taught by SK 2393 without the words "after a root". So it is not ārdhadhātuka, and takes no iṭ.
Rule. The prathama endings of luṭ are replaced, in order, by ḍā, rau and ras: singular, dual, plural.
ḍā is marked with ḍ; it is ḍit. Because of that, the ṭi of what precedes it is dropped by SK 316 (ṭeḥ). This happens even though the base is not bha. The ṭi is the last vowel with what follows it (SK 79).
Example.
bhū + i + tās + ti → ti is replaced by ḍā → bhū + i + tās + ā
→ guṇa of ū (SK 2168) and av (SK 61) → bhavitās + ā
→ ḍā is ḍit → the ṭi of tās, that is ās, is dropped → bhavit ā.
The next two rules finish the form.
Applies when / blocked by. Why does ḍā replace the whole ti? One of Bhaṭṭoji's commentators says that before it takes its place, ḍā is not yet an affix. So its ḍ is not yet an it, and ḍā counts as more than one sound (SK 45).
Watch out. SK 316 normally needs a bha base. Here it applies without one. Bhaṭṭoji says why: the ḍ-marker would otherwise serve no purpose (ḍittvasāmarthya).
Rule. Before a sārvadhātuka or ārdhadhātuka, guṇa replaces the ik of an aṅga that ends in the augment puk, or that has a light penultimate (laghūpadha). The ik is the aṅga's own vowel. The conditions come from SK 2168.
The ik is not directly before the affix: one consonant stands between. Bhaṭṭoji says the rule works across the one sound that must intervene (yena nāvyavadhānam). The very wording of the rule proves it, since a penultimate vowel always has one sound after it. It does not work across more than one sound.
Example.
bhavit ā → the aṅga is bhavit; its penultimate is the i of iṭ, a light ik
→ ā stands in place of ti, so it counts as sārvadhātuka (SK 49)
→ guṇa of that i is due by this rule.
SK 2190 stops it.
Across more than one sound: bhinatti "he splits" — the i of bhid is separated by more than one sound, so no guṇa.
Applies when / blocked by. Is bhavit an aṅga? The ending was added after bhū. One of Bhaṭṭoji's commentators answers with SK 199: what includes the vikaraṇa is also an aṅga.
Rule. Guṇa and vṛddhi do not come for the roots dīdhī and vevī, nor for the augment iṭ. The words come by anuvṛtti: "guṇa and vṛddhi" from SK 34, "not" from SK 2656.
Example.
bhavit ā → the i is the augment iṭ → no guṇa
→ bhavitā "he will be".
Applies when / blocked by. Which iṭ? A note in the printed edition says only the augment is meant, though it is named beside two roots.
Rule. tās and the root as "be" undergo lopa, deletion, when an affix beginning with s follows. "Beginning with s" is carried from SK 2342.
Example.
madhyama singular: bhavitās + si → si begins with s → the s of tās goes
→ bhavitāsi "you will be".
The rule names only "lopa of tās". The forms show that what goes is its final s.
Rule. tās is deleted in the same way when an affix beginning with r follows. as is not included here, since no affix beginning with r can come after it.
Example.
bhavitās + rau → the s goes → bhavitārau "the two will be".
bhavitās + ras → bhavitāras → bhavitāraḥ "they will be".
Bhaṭṭoji then lists the rest of luṭ:
bhavitāsi, bhavitāsthaḥ, bhavitāstha "you, you two, you all will be"
bhavitāsmi, bhavitāsvaḥ, bhavitāsmaḥ "I, we two, we will be".
Watch out. Before thas, tha, mi, vas, mas the s of tās stays: bhavitāsthaḥ, bhavitāsmi. Only an s- or r-initial affix removes it.
Rule. lṛṭ comes after a root in the future sense. It comes when no other action is being done for the sake of this future action (śeṣa, "the rest"). By ca, it also comes when one is. "The rest" means outside the scope of SK 3175, which gives other affixes when one action is done for the sake of another.
Example.
bhū + lṛṭ → bhū + ti
→ sya (SK 2186) → bhū + sya + ti
→ sya begins with a val → iṭ (SK 2184) → bhū + i + sya + ti
→ guṇa, av → bhavi + sya + ti
→ s of the affix after i becomes ṣ (SK 212) → bhaviṣyati "he will be".
The whole lṛṭ:
bhaviṣyati, bhaviṣyataḥ, bhaviṣyanti
bhaviṣyasi, bhaviṣyathaḥ, bhaviṣyatha
bhaviṣyāmi, bhaviṣyāvaḥ, bhaviṣyāmaḥ — lengthened by SK 2170.
Applies when / blocked by. Could "śeṣe ca" simply be dropped? Haradatta, an earlier commentator, thought so. Bhaṭṭoji's commentators both call this doubtful, and scholars discuss this.
bhava + ti → bhavatu; bhava + hi → bhava; bhava + ā + ni → bhavāniloṭ is used in the senses "injunction" and the rest, and in blessing. It is built on the same stem as the present, bhava, with śap. But its endings change. SK 2194–2195 give loṭ its senses. SK 2196 turns the final i of an ending into u: bhavatu "let him be". SK 2197 gives tāt in blessings: bhavatāt. SK 2198 makes loṭ borrow two operations of laṅ: tām and the others (SK 2199), and loss of s (SK 2200). SK 2201–2202 give hi for si, then delete it after a: bhava "be!". SK 2203–2204 give ni for mi, and add ā to the uttama: bhavāni "let me be".
Rule. loṭ also comes after a root in the senses vidhi (injunction) and the rest. The six senses are carried from SK 2208, where they are defined: vidhi, nimantraṇa, āmantraṇa, adhīṣṭa, saṃpraśna, prārthana. SK 2208 teaches them for liṅ, and comes in the next part of this chapter.
Example.
bhū, in one of these senses
→ bhū + loṭ → bhū + l → bhū + tip
→ śap, guṇa, av, as in the present → bhava + ti.
SK 2196 finishes the form.
Rule. liṅ and loṭ also come in the sense of benediction (āśīḥ). One of Bhaṭṭoji's commentators defines āśīḥ as the wish to obtain what is desired. Bhaṭṭoji gives the sūtra with no gloss.
Example.
bhū, in a blessing
→ bhū + loṭ → bhavatu or bhavatāt "may he be" (SK 2196, 2197).
Applies when / blocked by. Why not simply add "benediction" to the list of SK 2208? One of Bhaṭṭoji's commentators answers. The word āśiṣi is needed here anyway for the next sūtra of the Aṣṭādhyāyī (SK 3313). And as a separate, later rule, this one blocks rules like SK 2819 in its own field, rather than being blocked by them.
Rule. The i of a loṭ ending is replaced by u.
Example.
bhava + ti → the i of ti becomes u
→ bhavatu "let him be".
In the same way, bhava + anti (from jhi, SK 2169) → bhavantu "let them be".
Watch out. The i of hi (SK 2201) and of ni (SK 2203) is not changed. The reason is given at SK 2204. Neither of Bhaṭṭoji's commentators comments on this sūtra.
Rule. In benediction, tu and hi are optionally (anyatarasyām) replaced by tātaṅ. Its ṅ is it and its last a is for pronunciation, so the substitute is tāt. It replaces the whole of tu or hi.
Example.
bhavatu (in a blessing) → the whole tu is replaced by tāt
→ bhavatāt "may he be".
The other option keeps bhavatu.
Applies when / blocked by. Two rules pull different ways here. tāt has more than one sound, so by SK 45 it replaces the whole. But it is marked with ṅ, so by SK 43 it replaces only the last sound. SK 43 is normally an exception to SK 45, and an exception wins.
Bhaṭṭoji explains why SK 43 does not win here. It has already done its work in anaṅ and similar substitutes, whose ṅ has no other job. The ṅ of tātaṅ has other jobs: it stops guṇa and vṛddhi, and brings saṃprasāraṇa. So SK 43 applies here only "sluggishly". The later rule, SK 45, blocks it. In this field the general rule (utsarga) and the exception (apavāda) are of equal strength.
One of Bhaṭṭoji's commentators gives the jobs of the ṅ with forms. It stops guṇa in dviṣṭāt and vṛddhi in stutāt, yutāt. It brings saṃprasāraṇa in uṣṭāt. It stops an augment in brūtāt.
Rule. The operations taught in place of laṅ apply to loṭ as they do to laṅ. laṅ is the lakāra of the past not of today, taught at SK 2205. Bhaṭṭoji names the two operations meant: tām and the rest (SK 2199), and the loss of s (SK 2200).
Example.
bhava + tas (loṭ) → loṭ is treated like laṅ
→ tas becomes tām (SK 2199) → bhavatām "let the two be".
Applies when / blocked by. The augment aṭ of laṅ (SK 2206) is not carried over. Bhaṭṭoji's commentators both give the reason. Only what is taught in place of laṅ is carried over. aṭ is taught for the aṅga when laṅ follows, not in place of laṅ. So there is no a- before the loṭ forms.
Rule. For a ṅit lakāra, four endings are replaced, in order: tas → tām, thas → tam, tha → ta, mip → am. A ṅit lakāra is one marked with ṅ: laṅ, liṅ, luṅ, lṛṅ. The word "ṅit" is carried from SK 2200.
Example.
Through SK 2198, loṭ is treated like the ṅit laṅ:
bhava + tas → bhavatām "let the two be"
bhava + thas → bhavatam "be, you two!"
bhava + tha → bhavata "be, you all!".
Watch out. The fourth change, mip → am, does not reach loṭ. SK 2203 gives ni for its mi first, as one of Bhaṭṭoji's commentators explains.
Rule. An uttama ending of a ṅit lakāra that ends in s always loses it. By SK 42, only the final s goes. "Uttama", "ending in s" and "loss" are carried from SK 3428 and SK 3426, taught later. nitya "always" is said to stop an option carried from SK 3430.
Example.
Bhaṭṭoji lists two forms here, made by the rules around it:
bhavatām (tas → tām, SK 2199) and bhavantu (jhi → anti, SK 2169; i → u, SK 2196).
The rule itself acts in the uttama, at SK 2204:
bhavāvas → bhavāva "let us two be"; bhavāmas → bhavāma "let us be".
Applies when / blocked by. One of Bhaṭṭoji's commentators says the rule is placed here only incidentally. Its real work is in the uttama.
Watch out. Read "the uttama ending of a ṅit lakāra", not "the s of any ṅit uttama". One of Bhaṭṭoji's commentators warns that the second reading would wrongly reach bhūyāsam. The present bhavāvaḥ, bhavāmaḥ keeps its s, since laṭ is not ṅit.
Rule. The ending sip of loṭ, heard as si, is replaced by hi. And hi is not pit: it does not count as marked with p. "Of loṭ" is carried from SK 2198.
Example.
bhava + si (loṭ) → hi → bhava + hi.
SK 2202 finishes the form.
Applies when / blocked by. Why "not pit"? One of Bhaṭṭoji's commentators explains with stuhi "praise!". sip is pit; its substitute hi would count as pit too. Then SK 2234 would not treat hi as ṅit, and guṇa would come. Since hi is not pit, it is ṅit, and guṇa is barred: stuhi.
Rule. hi is deleted by luk when it comes after a. "luk" is carried from SK 2329.
Example.
bhava + hi → hi after a → luk
→ bhava "be!".
In benediction: bhava + hi → tāt (SK 2197) → bhavatāt "may you be".
Dual and plural: bhavatam, bhavata (SK 2199).
Applies when / blocked by. In benediction, which wins: luk of hi, or tāt for hi? tāt wins, since SK 2197 is later than this rule. Bhaṭṭoji's commentators and a note in the printed edition all say so.
Rule. The ending mip of loṭ, heard as mi, is replaced by ni. "Of loṭ" is carried from SK 2198.
Example.
bhava + mi (loṭ) → ni, not am → bhava + ni.
SK 2204 finishes the form.
Applies when / blocked by. One of Bhaṭṭoji's commentators says this rule comes in just as am (SK 2199) is about to apply, and pre-empts it.
Rule. An uttama ending of loṭ takes the augment āṭ, heard as ā. Being marked with ṭ, it goes at the front (SK 36). And it is treated as pit.
Example.
bhava + ni → bhava + ā + ni → a + ā merge (SK 85) → bhavāni "let me be".
bhava + vas → bhava + ā + vas → bhavāvas → s lost (SK 2200) → bhavāva "let us two be".
bhava + mas → bhavāmas → bhavāma "let us be".
The whole loṭ of bhū:
bhavatu / bhavatāt, bhavatām, bhavantu
bhava / bhavatāt, bhavatam, bhavata
bhavāni, bhavāva, bhavāma.
Applies when / blocked by. Why "pit"? One of Bhaṭṭoji's commentators explains with stavāni. Without it, the augmented ending would be ṅit by SK 2234, and guṇa would be barred. With ni, which replaces the pit mip, the augmented ending would count as pit anyway; the word matters for va and ma. Bhaṭṭoji's commentators differ on what "it" (saḥ) refers to, the augment or the ending; scholars discuss this.
Watch out. Why do hi and ni keep their i, against SK 2196? Bhaṭṭoji says it is because the i is deliberately uttered in them. One of his commentators spells it out. If the i were meant to change, the sūtras would simply have said hu and nu. By the same reasoning, the i of ni is not lost by SK 2207 (loss of i in a ṅit lakāra), carried over through SK 2198. Otherwise the sūtra would have said na.
bhava + ti → a + bhava + ti → abhavatThis group builds laṅ, the past tense used for a time before today. Three rules do the work. SK 2205 chooses the lakāra. SK 2206 puts the augment a in front of the verb stem. SK 2207 removes the final i of the ending. The running example starts from bhava + ti, the stem of bhū "to be" with its ending. It takes a in front and loses its i, giving abhavat "he was".
Rule. Laṅ comes after a root when the action is past and not of today. Anadyatana means "not belonging to today". Two headings are in force here: dhātoḥ "after a root" (SK 2829) and bhūte "in the past" (SK 2995). So Bhaṭṭoji glosses the rule: laṅ comes after a root that expresses a past sense not belonging to today.
Example.
bhū "to be", said of something that happened before today
→ laṅ is added after the root → bhū + laṅ
→ the ending, the vikaraṇa and the augment follow (SK 2206–2207) → abhavat "he was".
Bhaṭṭoji gives no forms under this rule. The forms come at SK 2207.
Applies when / blocked by. Anadyatana is a bahuvrīhi compound, one that names something outside its own members: "a time in which today has no place". So if the time spoken of takes in today, laṅ is not used. Only luṅ (SK 2218) is used then.
Watch out. Laṅ is not simply "the past". It is the past with today excluded. The general past is luṅ, taught at SK 2218.
Rule. When luṅ, laṅ or lṛṅ follows, the aṅga receives the augment aṭ, and aṭ is udātta. The aṅga is the base to which an affix is added; the heading aṅgasya (SK 200) is in force. An augment (āgama) is a sound added to something and made part of it. In aṭ the ṭ is a marker (it), so what is heard is a. It stands at the front of the aṅga.
Example.
bhū + laṅ → the ending tip replaces the l → bhū + ti
→ the vikaraṇa śap (SK 2167) → bhū + a + ti
→ guṇa of ū (SK 2168) and av for o → bhava + ti
→ this rule: the augment aṭ before the aṅga → a + bhava + ti.
SK 2207 then gives abhavat.
Applies when / blocked by. Why does aṭ come after śap, guṇa and av? One of Bhaṭṭoji's commentators explains that tip, śap, guṇa and av come first. Each wins by its own priority: it is inner (antaraṅga), or always applicable (nitya), or later in the order (para). So aṭ is added to the aṅga together with its vikaraṇa. Another view puts aṭ in while only the bare lakāra stands; scholars discuss this. Bhaṭṭoji himself says nothing on the order.
Watch out. Aṭ is taught for the aṅga when laṅ follows. It is not taught in place of laṅ. So SK 2198 (loṭo laṅvat), which lends loṭ the operations taught in place of laṅ, does not bring aṭ. That is why the imperative is bhavatu, with no a in front.
Rule. In a lakāra marked with ṅ (a ṅit lakāra), a parasmaipada ending that ends in i loses its i. The words "loss" (lopaḥ) and "parasmaipada" are carried forward from SK 3426. The word ṅitaḥ "of a ṅit lakāra" comes from SK 2200. By SK 42 only the final sound is lost. Laṅ, liṅ, luṅ and lṛṅ are the ṅit lakāras.
Example.
a + bhava + ti → the ending ti ends in i and belongs to laṅ → the i is lost
→ abhavat "he was".
The full paradigm of bhū in laṅ:
abhavat, abhavatām, abhavan (third person: one, two, many)abhavaḥ, abhavatam, abhavata (second person)abhavam, abhavāva, abhavāma (first person).One of Bhaṭṭoji's commentators derives each form. In abhavan the jh of jhi becomes ant (SK 2169), the i is lost by this rule, and the final t of the conjunct nt is lost. abhavatām, abhavatam, abhavata and abhavam have tām, tam, ta and am by SK 2199. In abhavaḥ the i of si is lost, and the s becomes visarga. In abhavāva and abhavāma the s of vas, mas is lost by SK 2200, and the a is lengthened by SK 2170.
Watch out. The rule removes the i of an ending that ends in i. It does not remove every i found in a parasmaipada ending. One of Bhaṭṭoji's commentators says that an earlier wording, "loss of i in the parasmaipada", goes too far. It would wrongly reach forms like bhavet, where an i stands inside the substitute iy (SK 2212).
bhava + yās + s + t → bhava + yā + t → bhava + iy + t → bhavetThis group builds vidhiliṅ, the liṅ of command, invitation, request and the like. SK 2208 gives its senses. Two augments come next: yās before the parasmaipada ending (SK 2209) and s before a t or th (SK 2210). Then the s's are lost (SK 2211), and yā becomes iy after a (SK 2212). The third person plural takes us (SK 2213–2214). The running example: bhava + yās + s + t loses both s's, turns yā into iy, and ends as bhavet "he should be".
Rule. Liṅ comes after a root in six senses. Bhaṭṭoji says the sense is either made known (dyotya) or expressed (vācya) by the liṅ. He keeps both views; scholars discuss this. The six senses:
Bhaṭṭoji defines the first four. Bhaṭṭoji's commentators define the last two.
Example. One of Bhaṭṭoji's commentators gives a sentence for most of the senses:
bhavān vastraṃ kṣālayet "your honour should wash the cloth" (a master to a servant);iha śrāddhe bhavān bhuñjīta "your honour should eat at this śrāddha";māṇavakaṃ bhavān adhyāpayet "your honour should teach the boy";kiṃ bhoḥ vyākaraṇaṃ bhavān adhīyīta "sir, should your honour study grammar?";bhavān annaṃ me dadyāt "may your honour give me food".For bhū the steps begin:
bhū + liṅ → bhū + tip → śap, guṇa, av → bhava + ti.
The next rules turn this into bhavet.
Applies when / blocked by. Bhaṭṭoji remarks that one could simply say "liṅ in prompting (pravartanā)". The first four senses are all kinds of prompting. They are listed one by one for the sake of detail.
Rule. The parasmaipada endings of liṅ receive the augment yāsuṭ. It is udātta, and it is ṅit, marked with ṅ. The word liṅaḥ "of liṅ" is carried forward from SK 2255. In yāsuṭ the ṭ is a marker and the u is only for pronunciation. What is heard is yās. Because it is ṭ-marked, it stands at the front of the ending.
Example.
bhava + ti → the i is lost by SK 2207 → bhava + t
→ this rule: yās at the front of the ending → bhava + yās + t.
SK 2210 adds a second s.
Applies when / blocked by. Why state that yāsuṭ is ṅit? One of Bhaṭṭoji's commentators gives the practical reason: it stops guṇa in stuyāt "may he praise". Bhaṭṭoji draws a lesson from the statement. SK 49 treats a substitute like the thing it replaces, but not in an operation that depends on a single letter. The ṅ of liṅ is such a single letter. Without the exclusion, yāsuṭ, being part of the liṅ's substitute, would already be ṅit by SK 49, and saying "ṅit" would be pointless. So the explicit "ṅit" shows that this exclusion sometimes holds even for marker letters. Bhaṭṭoji adds a second sign of the same thing: the substitute śānac is given its own ś marker (SK 2557).
Watch out. Yāsuṭ is for the parasmaipada endings only. The ātmanepada endings of liṅ take a different augment, sīyuṭ (SK 2255).
Rule. The t and the th of liṅ receive the augment suṭ. In the sūtra ti means just t; the i is for pronunciation. In suṭ too the ṭ is a marker and the u is for pronunciation. What is heard is s, placed in front of the t or th.
Example.
bhava + yās + t → this rule: s before the t → bhava + yās + s + t.
SK 2211 removes both s's.
Applies when / blocked by. Suṭ is the later rule. Why does it not block yāsuṭ? Bhaṭṭoji says their domains differ. Yāsuṭ is an augment of the liṅ ending. Suṭ is an augment of the t or th within it. Both can apply together, so neither blocks the other.
Rule. In a sārvadhātuka liṅ, an s that is not final is lost. The word sārvadhātuke is carried forward from SK 2474. So the rule works only where the liṅ ending counts as sārvadhātuka.
Example.
bhava + yās + s + t → both s's are non-final → both are lost
→ bhava + yā + t.
SK 2212 continues.
Bhaṭṭoji says "both s's go". One of Bhaṭṭoji's commentators explains why the s of suṭ counts as part of the liṅ. Suṭ is part of the ending, and the ending stands for the liṅ. A part of a part is a part of the whole. Whether the two s's go at once or one after the other, scholars discuss this.
Applies when / blocked by. The s of suṭ is heard in the benedictive liṅ (āśīrliṅ), as in bhūyāstām "may the two be". There the liṅ is not sārvadhātuka (SK 2215), so this rule does not apply. Bhaṭṭoji adds that it is heard still more clearly in the ātmanepada, as in edhiṣīṣṭa (SK 2257).
Watch out. One of Bhaṭṭoji's commentators notes that in bhava + yās + s + t another rule, SK 380, would also remove the s's. This rule is still needed where no such cluster arises. In edha + sīy + ta (SK 2255) the s stands before the vowel ī, so SK 380 cannot remove it. This rule does, giving edheta.
Rule. When yā is part of a sārvadhātuka and comes after a short a, it is replaced by iy. "Short a" renders at, the a spoken with t after it.
Example.
bhava + yā + t → yā after short a → iy → bhava + iy + t
→ a + i gives e by SK 69 (ādguṇaḥ) → bhavey + t
→ y is lost before a consonant by SK 873 (lopo vyor vali) → bhavet "he should be".
In the same way: bhavetām "the two should be".
Applies when / blocked by. Why does the rule say "of a sārvadhātuka"? Bhaṭṭoji answers with cikīrṣyāt "may he wish to do". That is a benedictive liṅ, which is not sārvadhātuka. Without the word, iy would come there too, giving the wrong *\cikīrṣet.
Bhaṭṭoji also shows why the word is needed in bhavet itself. He uses a maxim: an exception standing in the middle blocks only the rules before it, not those after it. Without the word, this rule (A 7.2.80) would block only SK 2308 (ato lopaḥ, A 6.4.48), which stands before it in Pāṇini's order. In bhava + yā + t, the lengthening rule SK 2170 (A 7.3.101), which stands after it, would then take effect. The a would become ā, and iy would have no a to follow.
Watch out. Iy replaces yā, not just the ā. Then guṇa and the loss of y finish the form.
Rule. The ending jhi of liṅ is replaced by jus. The j is a marker, so what is heard is us. Being more than one sound, jus replaces the whole of jhi. It comes where SK 2169 would otherwise give ant.
Example.
bhava + jhi → this rule → bhava + us
→ with yāsuṭ (SK 2209) → bhava + yās + us.
SK 2214 shows how this becomes bhaveyuḥ.
Rule. When the u of us follows an a-vowel that does not end a pada, the two are replaced by one sound, the later one. This single replacement is pararūpa, "the form of the latter". The heading "one for both" (SK 68) is in force; aci (SK 47), āt (SK 69) and pararūpam (SK 78) are carried forward.
Example.
The rule's own case, from one of Bhaṭṭoji's commentators:
bhind + yā + us → ā + u → u → bhindyuḥ "they should split".
The case Bhaṭṭoji discusses:
bhava + yās + us → iy replaces yās (SK 2212, read as below) → bhava + iy + us
→ a + i gives e by SK 69 → bhaveyus → bhaveyuḥ "they should be".
The rest of the paradigm: bhaveḥ, bhavetam, bhaveta; bhaveyam, bhaveva, bhavema.
Applies when / blocked by. In bhava + yā + us, this rule and SK 2212 compete. The earlier grammarians (prāñcaḥ) said SK 2212 wins, because it is later and because it is nitya. Bhaṭṭoji disagrees. He says pararūpa is antaraṅga, the "inner" operation, so it would properly come first. It would turn ā + u into u, and yā would be gone.
Bhaṭṭoji's answer is to read SK 2212 differently. After a, iy replaces yās, with its s. So iy comes before the s is lost. It is an exception to SK 2211. While the s stands, the u of us does not follow an a-vowel, so pararūpa cannot apply. On this reading the sūtra should sound ato yā iyaḥ. Bhaṭṭoji calls its actual sandhi yeyaḥ a seer's usage (ārṣa).
Watch out. The order matters. Iy must replace yās while its s still stands. If the s were lost first, pararūpa would swallow the ā, and iy could no longer come.
bhū + yās + s + t → bhūyātThis group builds āśīrliṅ, the liṅ of blessing. SK 2215 makes its endings ārdhadhātuka only. So there is no śap, and the rules of SK 2211–2212 do not reach it. SK 2216 makes its yās kit. SK 2217 says a kit affix causes no guṇa. The running example: bhū + yās + s + t loses its two s's and keeps the root vowel ū, giving bhūyāt "may he be".
Rule. In benediction (āśīḥ), the tiṅ ending that replaces liṅ is named ārdhadhātuka only. The word "only" (eva) is carried forward from SK 2463. Benediction is a blessing, a wish for someone's good.
Example.
bhū + liṅ (benediction) → bhū + ti
→ the ending is ārdhadhātuka, not sārvadhātuka → no śap (SK 2167 needs a sārvadhātuka)
→ the base stays bhū.
SK 2216–2217 finish the form bhūyāt.
Applies when / blocked by. Why "only"? One of Bhaṭṭoji's commentators answers. A tiṅ ending is sārvadhātuka by SK 2166. Without "only", the ending would carry both names. Then śap and the other vikaraṇas would come optionally.
Watch out. Since this liṅ is not sārvadhātuka, SK 2211 (loss of s) and SK 2212 (iy) do not apply to it. That is why the benedictive keeps an s in bhūyāstām.
Rule. In benediction, the yāsuṭ of liṅ is kit, marked as if with k. This is an exception to SK 2209, which made yāsuṭ ṅit.
Example. One of Bhaṭṭoji's commentators gives the steps:
bhū + ti → the i is lost (SK 2207) → bhū + t
→ yāsuṭ (SK 2209) → bhū + yās + t
→ no iy: yās does not follow a, and the ending is not sārvadhātuka
→ suṭ (SK 2210) → bhū + yās + s + t
→ SK 380 (skoḥ saṃyogādyor ante ca) removes both s's → bhūyāt "may he be".
SK 380 removes an s or k that begins a consonant cluster at the end of a pada. Each s here is such a sound.
Applies when / blocked by. A ṅit affix already blocks guṇa. So why make yāsuṭ kit? One of Bhaṭṭoji's commentators says kit does work that ṅit cannot. It brings saṃprasāraṇa in ijyāt "may he sacrifice" (SK 2409). Saṃprasāraṇa is the change of y, v, r into the vowels i, u, ṛ: yaj becomes ij. The other adds jāgaryāt "may he be awake": a ṅit affix would forbid its guṇa, but a kit one does not.
Watch out. Which clause of SK 380 removes the s of yās in bhūyāstām, bhūyāstam and bhūyāsta? Scholars discuss this.
Rule. Guṇa and vṛddhi taught by naming ik do not occur when their cause is git, kit or ṅit: marked with g, k or ṅ. The word "not" (na) is carried forward from SK 2656. The vowels i, u, ṛ, ḷ are the ik vowels. The g is read into the sūtra by praśleṣa, "merging": it is hidden in the k.
Example.
bhū + yās + s + t → yāsuṭ is kit (SK 2216) → no guṇa of ū (SK 2168 is blocked)
→ bhūyāt, not *\bhoyāt.
The full benedictive paradigm of bhū:
bhūyāt, bhūyāstām, bhūyāsuḥ (third person: one, two, many)bhūyāḥ, bhūyāstam, bhūyāsta (second person)bhūyāsam, bhūyāsva, bhūyāsma (first person).In bhūyāsuḥ the ending is us (SK 2213). In bhūyāsva, bhūyāsma the final s of vas, mas is lost by SK 2200.
Applies when / blocked by. Bhaṭṭoji's commentators explain three details of the wording:
jiṣṇuḥ "victorious", made with the affix gsnu. Without "git", guṇa would come there.chinnam "cut" and bhinnam "split", a consonant stands between the vowel and the affix. The ban must still hold.laigavāyanaḥ, whose vṛddhi is taught otherwise, is not affected.Watch out. The ban is on guṇa and vṛddhi. Other operations caused by a kit affix, such as saṃprasāraṇa, still take place.
bhū + luṅ → abhūt (SK 2221 ff.)One rule opens the aorist, luṅ. It gives only the lakāra. The forms of bhū in luṅ, starting with abhūt "he was", follow in group II.A.
Rule. Luṅ comes after a root that expresses a past sense. The headings dhātoḥ (SK 2829) and bhūte (SK 2995) are in force. Unlike laṅ (SK 2205), luṅ has no restriction "not of today".
Example.
bhū "to be", past → luṅ → bhū + luṅ
→ cli, sic and the loss of sic follow (SK 2221–2223) → abhūt "he was".
Bhaṭṭoji gives no forms here.
Watch out. Where the time spoken of includes today, only luṅ is used, not laṅ (see SK 2205).
bhū + luṅ → a-bhū-s-t → abhūt; abhūvan; mā bhavān bhūt; abhaviṣyatThis group finishes bhū. First come two rules for the particle mā (SK 2219–2220). Then the aorist is built. Cli comes (SK 2221), and sic replaces it (SK 2222). After bhū, sic is lost (SK 2223). There is no guṇa (SK 2224), and no ī (SK 2225). The third person plural follows (SK 2226–2227). SK 2228 drops the augment after mā. SK 2229 gives lṛṅ, the conditional. The running example: a-bhū-s-t loses its s and gives abhūt "he was". The plural is abhūvan. With mā the a goes: mā bhavān bhūt. The conditional is abhaviṣyat.
Rule. When māṅ is used, luṅ comes after the root. Māṅ is the particle mā, cited with the marker ṅ. Bhaṭṭoji calls this rule an exception to all the lakāras (sarvalakārāpavāda).
Example.
mā + bhū → luṅ, whatever the time → bhū + luṅ
→ the aorist form without its augment (SK 2228) → mā bhavān bhūt "may your honour not be".
Applies when / blocked by. Why is this rule needed, when luṅ (SK 2218) already gives luṅ? One of Bhaṭṭoji's commentators answers. SK 2218 gives luṅ only for the past. This rule sets aside every other lakāra when māṅ is used.
Watch out. In mā 'stu "let it not be", the verb is not luṅ. Bhaṭṭoji's commentators say that this mā is a different indeclinable, meaning prohibition. It is not māṅ.
Rule. When māṅ has sma after it, laṅ comes. By the word ca "and", luṅ comes too. Ca draws the preceding rule, māṅi luṅ, into this one.
Example.
mā sma + bhū → laṅ → mā sma bhavat
or → luṅ → mā sma bhūt.
Both forms are given at SK 2228, where the augment is dropped.
Applies when / blocked by. One of Bhaṭṭoji's commentators notes that this too is an exception to all the lakāras.
Rule. When luṅ follows, the root takes the affix cli. The i of cli is only for pronunciation. Bhaṭṭoji calls it an exception to śap and the rest. That is, cli displaces the vikaraṇa each root class would take: śap, śyan and so on.
Example.
bhū + luṅ → bhū + tip → the i is lost (SK 2207) → bhū + t
→ this rule: cli after the root → bhū + cli + t.
Watch out. Cli itself is not heard in the finished form. For bhū, SK 2222 replaces it with sic.
Rule. Sic replaces cli. In sic the i and the c are markers. What is heard is s.
Example.
bhū + cli + t → sic for cli → bhū + s + t.
With the augment aṭ (SK 2206) this is a-bhū-s-t.
Applies when / blocked by. Bhaṭṭoji's commentators give the use of each marker. The c brings the accent of a cit affix (one marked with c). The i stops the loss of the penultimate nasal in amaṃsta "he thought" (SK 415).
Rule. After gā, sthā, the ghu roots, pā and bhū, sic undergoes luk before a parasmaipada ending. Luk is total disappearance of an affix. The word luk is carried forward from SK 1276. The ghu roots are dā and dhā.
Example.
a-bhū-s-t → sic after bhū, before the parasmaipada t → luk of sic
→ a-bhū-t.
Whether ū now takes guṇa is settled by SK 2224.
Applies when / blocked by. Bhaṭṭoji says which gā and which pā are meant. Gā is the substitute of iṇ "to go" (SK 2458). Pā is the root "to drink", which takes the substitute piba.
Watch out. SK 1276, the source of luk, is separated from this rule by other rules. A rule with the word ślu (SK 2489) is not. Even so, Bhaṭṭoji's commentators say only luk is carried here, not ślu. They rely on the accepted explanation (vyākhyāna).
Rule. Bhū and sū take no guṇa when a sārvadhātuka tiṅ ending follows. "Guṇa" comes from SK 2346. "Not" and "sārvadhātuka" come from SK 2503.
Example.
a-bhū-t → the t is a tiṅ ending and sārvadhātuka, and it follows bhū directly
→ guṇa by SK 2168 is blocked → abhūt "he was", not *\abhot.
Applies when / blocked by. Bhaṭṭoji's commentators say which sū is meant. It is ṣūṅ "to give birth", of the class whose vikaraṇa is lost. The sū roots that take śa or śyan are not meant. There the vikaraṇa stands between root and ending, so the rule has nothing to do.
Rule. An apṛkta consonant ending takes the augment īṭ when it follows a sic or an as that is still present. An apṛkta affix is one that consists of a single sound. Īṭ is heard as ī. It is carried from SK 2452, and "consonant" (hal) from SK 2443.
Bhaṭṭoji analyses the compound astisicaḥ. Sic and as first form a dvandva, a compound of words joined by "and": sicas. This joins the indeclinable asti "present, existing" in a karmadhāraya, a compound whose members name one and the same thing: "a sic or as that exists". The c of sic does not change inside the compound; the sūtra's own usage treats it so.
Example.
a-bhū-t → the t is apṛkta, but the sic has gone by luk (SK 2223)
→ no īṭ → abhūt.
abhūtām "the two were" follows in the same way.
Applies when / blocked by. Why "consonant"? Bhaṭṭoji gives aidhiṣi "I grew". There the ending after sic is i, a vowel. Why "apṛkta"? aidhiṣṭa "he grew". There the ending ta after sic has two sounds.
Watch out. "Present" is the key word. Bhaṭṭoji's commentators reject an older analysis, "after as and after sic". An elided sic would still count for it. Then abhūt, agāt "he went" and asthāt "he stood" would all wrongly take ī.
Rule. After sic, after an abhyasta and after vid, the jhi of a ṅit lakāra is replaced by jus. An abhyasta is a doubled root. "Jhi → jus" is carried from SK 2213. "Ṅit" is carried from SK 2200.
Example.
a-bhū + jhi → the sic has gone, but an elided affix still counts (pratyayalakṣaṇa)
→ this rule would give jus here.
Bhaṭṭoji stops at this point: "when this is due…". SK 2227 answers.
Watch out. Left alone, this rule would wrongly give jus in the third person plural of bhū. The next rule prevents that.
Rule. When sic has undergone luk, jus replaces jhi only after an aṅga ending in ā. This is a restriction (niyama). SK 2226 already gives jus after sic. This rule limits it, after an elided sic, to ā-final roots.
Example.
a-bhū + jhi → bhū does not end in ā → no jus
→ jh becomes ant (SK 2169) → a-bhū + anti
→ the i is lost (SK 2207) → a-bhū + ant
→ the final t of the cluster nt is lost → a-bhū + an
→ the augment vuk (SK 2174) → abhūvan "they were".
The full aorist of bhū:
abhūt, abhūtām, abhūvan (third person: one, two, many)abhūḥ, abhūtam, abhūta (second person)abhūvam, abhūva, abhūma (first person).Applies when / blocked by. The sūtra is one word, ātaḥ "after ā". One of Bhaṭṭoji's commentators reads the sic condition in by carrying it forward from SK 2226. The other finds it in a vārttika, a supplementary statement on a sūtra: "after an ā-final root whose sic has gone". They agree on the result.
Watch out. abhūvam, abhūva and abhūma keep ū too. One of Bhaṭṭoji's commentators derives them with am for mip (SK 2199), and with vas, mas losing their s (SK 2200). In abhūvam the augment vuk comes before the vowel of am.
Rule. The augments aṭ and āṭ do not come when the verb is construed with māṅ. They are carried from SK 2206 and SK 2254.
Example.
mā + abhūt → no aṭ → mā bhavān bhūt "may your honour not be".
mā sma + abhavat (laṅ, SK 2220) → no aṭ → mā sma bhavat
or mā sma + abhūt (luṅ) → mā sma bhūt.
One of Bhaṭṭoji's commentators adds the plurals mā sma bhavan and mā sma bhūvan.
Watch out. Bhaṭṭoji gives examples of the missing aṭ only. The missing āṭ concerns roots that begin with a vowel (SK 2254).
Rule. Where a cause for liṅ is present, lṛṅ comes in the sense of the future. It comes when the action is understood as not accomplished (kriyātipatti). A cause for liṅ is, for example, a relation of cause and effect (hetuhetumadbhāva) between two clauses. "Future" is carried from SK 2794.
Example. One of Bhaṭṭoji's commentators gives the sentence suvṛṣṭiś ced abhaviṣyat tadā subhikṣam abhaviṣyat. It means "had there been good rain, there would have been plenty".
The steps for abhaviṣyat:
bhū + lṛṅ → bhū + tip → the i is lost (SK 2207) → bhū + t
→ sya in place of śap (SK 2186) → bhū + sya + t
→ the augment aṭ (SK 2206) → a + bhū + sya + t
→ iṭ, since sya begins with a consonant of the val group, every consonant but y (SK 2184) → a + bhū + i + sya + t
→ guṇa and av → a + bhav + i + sya + t
→ s becomes ṣ after i → abhaviṣyat.
The paradigm:
abhaviṣyat, abhaviṣyatām, abhaviṣyanabhaviṣyaḥ, abhaviṣyatam, abhaviṣyataabhaviṣyam, abhaviṣyāva, abhaviṣyāma.Applies when / blocked by. Bhaṭṭoji writes "cause and effect and the like". What "and the like" takes in, scholars discuss this. Bhaṭṭoji's commentators leave the causes of liṅ to the later section on the meanings of the lakāras.
pra + bhavāni → prabhavāṇi; pra-ni-bhavati → praṇibhavati / pranibhavatiThree rules close the treatment of bhū. SK 2230 puts the prefixes before the root. SK 2231 turns the n of āni into ṇ after a prefix with r. SK 2232 does the same, optionally, for the n of ni. Then the second root, edh "to grow", is introduced. The running example: pra + bhavāni gives prabhavāṇi "let me arise". Pra + ni + bhavati gives praṇibhavati or pranibhavati.
Rule. The words named gati and upasarga are used only before the root. An upasarga is a prefix such as pra or sam when joined to a verb (SK 22). Gati is a related name for such words (SK 23). They may not stand after the root, and may not be separated from it.
Example.
bhavāni "let me be" + pra → pra must stand before the root → pra + bhavāni.
SK 2231 turns this into prabhavāṇi.
Applies when / blocked by. Bhaṭṭoji's commentators point to two Vedic rules, SK 3391–3392. In the Veda a prefix may follow the root, or be separated from it. Those rules are needed precisely because this rule forbids both.
Rule. The n of āni, the loṭ ending, becomes ṇ after an r or ṣ that stands in an upasarga. The first person singular of loṭ ends in āni (SK 2203–2204). This change of n to ṇ is called ṇatva. "N becomes ṇ" is carried from SK 235, and "after an upasarga" from SK 2287.
Example.
pra + bhavāni → the r of pra stands in an upasarga
→ the n of āni → ṇ → prabhavāṇi "let me arise".
The sounds bh, a, v, ā stand between r and n. That is allowed by SK 197, which holds through the whole ṇatva section.
Applies when / blocked by. One of Bhaṭṭoji's commentators says why the rule is needed. The upasarga and the verb are two separate padas. The general ṇatva rule needs both sounds in one pada.
Two vārttikas follow.
duḥsthitiḥ "a bad state" keeps its s, and durbhavāni (loṭ of bhū with dur) keeps its n.antardhā "concealment" (with aṅ), antardhiḥ "concealment" (with ki), and antarbhavāṇi (loṭ of bhū with antar).Watch out. Dur has an r, but gives no ṇ. Antar is not in the list of prefixes, yet gives ṇ. Both are set by the vārttikas.
Rule. The n of the prefix ni, after an upasarga with r or ṣ, optionally becomes ṇ. This holds before any root other than gad, nad and the rest, which have their own rule (SK 2285). Roots that begin with k or kh, or end in ṣ, are excluded. Their shape is judged in upadeśa, the form in which the root is first taught.
Example.
pra + ni + bhavati → the r of pra, then the n of ni, before bhū
→ optional ṇ → praṇibhavati or pranibhavati.
Applies when / blocked by. Why must the option be stated? Bhaṭṭoji explains. Between a prefix and a root, joining in sound (saṃhitā) is obligatory, even when they are not compounded. He quotes a verse of Hari:
> Saṃhitā is obligatory within a single word, obligatory between root and upasarga, obligatory in a compound; but in a sentence it depends on what the speaker intends.
So the option cannot come from saṃhitā being optional. The sūtra must state it.
Why "in upadeśa"? One of Bhaṭṭoji's commentators gives pranicakāra. Its k has become c after doubling, yet the ban still holds, since the root began with k when taught.
Bhaṭṭoji then makes two points about meaning. First, the meanings in the list of roots, such as bhū "existence", are only examples (upalakṣaṇa). Bhū also means "arise", as in yāgāt svargo bhavati "heaven comes about from sacrifice". Second, an upasarga makes a particular meaning manifest (dyotaka); it does not itself carry it. He cites prabhavati, parābhavati, saṃbhavati, anubhavati, abhibhavati, udbhavati and paribhavati. One of Bhaṭṭoji's commentators glosses the related nouns. Parābhava is "defeat", saṃbhava "possibility", anubhava "enjoyment", abhibhava "injury", paribhava "contempt". Bhaṭṭoji quotes a second verse of Hari:
> By an upasarga the meaning of a root is forcibly understood as another, as in prahāra "striking", āhāra "eating", saṃhāra "destroying", vihāra "sporting", parihāra "avoiding".
After this sūtra Bhaṭṭoji turns to the second root of the list, edha "to grow" (vṛddhau). It opens a run of thirty-six roots, up to kattha "to boast". All have an anudātta vowel as their marker (anudāttet). By SK 2157 they take only ātmanepada endings. The next groups build edh through every lakāra.
edh + śap + ta → edha + ta → edhateEdh "to grow" takes only ātmanepada endings. This group builds its present tense, laṭ. SK 2233 turns the last vowel of each ending into e. SK 2234 makes endings like ātām count as ṅit. SK 2235 then turns their ā into iy. SK 2236 replaces thās with se. The running example: edh + śap + ta gives edha + ta, and the a of ta becomes e: edhate "he grows".
Rule. In a lakāra marked with ṭ, such as laṭ, the ṭi of each ātmanepada ending becomes e. The ṭi of a unit is its last vowel with whatever follows it. The heading lasya (SK 2153) is in force.
Example.
edh + laṭ → the ātmanepada third person singular ending → edh + ta
→ śap (SK 2167) → edha + ta
→ the ṭi of ta is its a → e → edhate "he grows".
One of Bhaṭṭoji's commentators notes that in ta the single a is both the last vowel and all that follows it, so it is the ṭi.
Applies when / blocked by. "Ātmanepada" here means the tiṅ endings of the ātmanepada only (the taṅ set). It does not include the participle affix āna. So pacamānaḥ "cooking" and yajamānaḥ "sacrificing" keep their a. Bhaṭṭoji's commentators both say so.
Watch out. The rule needs a ṭ-marked lakāra. Laṅ has no ṭ, so the laṅ endings of edh keep their vowel: aidhata (SK 2254), not *\aidhate.
Rule. A sārvadhātuka affix that is not pit (not marked with p) is treated as ṅit (ṅidvat). "Ṅit" is carried from SK 2461. The endings tip, sip and mip carry the marker p; the others do not.
Example.
edh + ātām (third person dual) → śap → edha + ātām
→ the ṭi ām becomes e (SK 2233) → edha + āte
→ āte is sārvadhātuka and not pit → treated as ṅit.
SK 2235 now acts on it.
Applies when / blocked by. One of Bhaṭṭoji's commentators explains "treated as". A non-ṅit cannot really be ṅit. So the rule makes it "like a ṅit".
Watch out. The rule is for sārvadhātuka affixes only. The liṭ endings are ārdhadhātuka (SK 2172), so this rule does not make them ṅit. Liṭ has its own rule (SK 2242).
Rule. After a short a, the ā of a ṅit affix is replaced by iy. "After short a" and "iy" are carried from SK 2212.
Example.
edha + āte → āte counts as ṅit (SK 2234) → its ā after short a → iy
→ edha + iy + te → a + i gives e (SK 69) → edhey + te
→ y is lost before a consonant (SK 873) → edhete "the two grow".
The plural:
edh + jha → śap → edha + jha → jh becomes ant (SK 2169) → edha + anta
→ a + a gives the later a (pararūpa) → edhanta → the ṭi a becomes e → edhante "they grow".
Watch out. Only the ā is replaced, not the whole ending. The te of āte stays.
Rule. In a ṭ-marked lakāra, the ending thās is replaced by se. "Ṭ-marked" is carried from SK 2233.
Example.
edha + thās (second person singular) → se → edhase "you grow".
The second person dual and plural: edhethe (from āthām, as edhete) and edhadhve (from dhvam).
The first person singular:
edha + iṭ → the ṭi of iṭ becomes e (SK 2233) → edha + e
→ vṛddhi (a + e → ai) would come
→ SK 191 (ato guṇe) gives the later vowel instead → edhe "I grow".
Also edhāvahe, edhāmahe, with the a lengthened before v, m (SK 2170).
The full laṭ: edhate, edhete, edhante; edhase, edhethe, edhadhve; edhe, edhāvahe, edhāmahe.
Watch out. In edhe, a + e does not give ai. SK 191 blocks the vṛddhi here, and the e alone remains.
edh + liṭ → edh-ām + [luk of liṭ] + kṛ-liṭ → edhāṃcakreEdh does not double in the perfect, liṭ. Instead it takes the affix ām, and a helping verb in liṭ follows. SK 2237 gives ām. SK 2238 removes the liṭ after ām. SK 2239 brings the helping verbs kṛ, bhū and as. SK 2240 gives kṛ the same pada as edh. The running example: edh + liṭ becomes edhām, and kṛ in liṭ follows, giving edhāṃcakre "he grew".
Rule. A root that begins with an ic vowel and contains a heavy (guru) vowel takes ām in liṭ. The root ṛcch is excepted. The ic vowels are all the vowels except a and ā. "Root" is carried from SK 2629, and "ām in liṭ" from SK 2306.
Example.
edh + liṭ → edh begins with e, an ic vowel; e is long, so heavy
→ ām after the root → edh + ām + liṭ.
SK 2238 continues.
Applies when / blocked by. One of Bhaṭṭoji's commentators gives the counter-examples. Why "heavy"? iyeṣa "he wished": iṣ has no heavy vowel, so it doubles instead. Why "not ṛcch"? Ṛcch begins with ṛ, an ic vowel, and its ṛ is heavy because the cluster cch follows. So it would take ām. The exception makes it double instead: ānarccha.
Watch out. The m of ām is not a marker. Bhaṭṭoji proves it from the roots ās and kās, which also take ām (SK 2324, SK 2306). If m were a marker, ām would go after the root's last vowel (SK 37). Then ā + ā would merge, leaving just ās, kās. Teaching ām for them would be pointless. So the m is heard: edhām.
Rule. After ām, the l-affix undergoes luk. So the liṭ that caused ām disappears. "Luk" is carried from SK 1276.
Example.
edh + ām + liṭ → luk of the liṭ after ām → edhām.
SK 2239 adds the helping verb, which carries the liṭ.
Applies when / blocked by. Bhaṭṭoji glosses "the l-affix after ām". Whether that word must be carried into the sūtra, scholars discuss this. Bhaṭṭoji's commentators agree that nothing but the l-affix is left for the rule to remove.
Watch out. One of Bhaṭṭoji's commentators adds that edhām counts as a pada, a finished word. It is treated as ending in a case-ending that has itself been removed.
Rule. After a form ending in ām, kṛ, bhū and as, followed by liṭ, are applied after it. This is anuprayoga, "after-application". Bhaṭṭoji's commentators add that the helping verb comes immediately after. It may not come before, and nothing may stand between.
Example.
edhām → kṛ "to do" with liṭ is applied after → edhām + kṛ + liṭ.
SK 2240 fixes its pada; SK 2241–2245 make edhāṃcakre.
Applies when / blocked by. How does kṛñ mean three roots? Bhaṭṭoji reasons from SK 2240. That rule names kṛñ to mark it off from other helping verbs. If kṛ were the only one, naming it there would be pointless. So kṛñ here is a pratyāhāra, a short name for a series. It runs from kṛ in SK 2117 (kṛbhvastiyoge) to the ñ of SK 2129. That takes in kṛ, bhū and as.
Sampadi "to become" also falls inside that series. It is not used, Bhaṭṭoji says, because its meaning cannot be joined with the base. The three helping verbs mean action in general. The base of ām means a particular action. The two meanings are construed as one and the same (abhedānvaya): "an action, namely growing".
Watch out. Nothing may stand between the ām form and its helping verb. Bhaṭṭoji's commentators call a usage like taṃ pātayāṃ prathamam āsa papāta paścāt an error. There prathamam stands between pātayām and āsa.
Rule. The after-applied kṛñ takes ātmanepada in the same way as the base of ām does. "Ātmanepada" is carried from SK 2157.
Bhaṭṭoji explains āmpratyaya. It is a bahuvrīhi: "that after which the affix ām comes". It is of the kind that leaves out its own members (atadguṇasaṃvijñāna). So it denotes the root, such as edh, without the ām.
Example.
edhām + kṛ + liṭ → edh takes ātmanepada (SK 2157) → so does kṛ
→ edhām + kṛ + ta.
Bhaṭṭoji leads into the rule at the end of SK 2239. For kṛ, when the fruit of the action goes to another, parasmaipada would be due. This rule gives ātmanepada even then.
Applies when / blocked by. The word pūrvavat "as with the earlier root" is carried from SK 2734. It forms a second sentence (vākyabheda), a restriction: ātmanepada only as with the base, never the opposite. So in indāṃcakāra "he was powerful" there is no ātmanepada, even when the fruit goes to the agent. Its base, idi "to be powerful", takes parasmaipada.
Watch out. The rule names kṛñ alone. One of Bhaṭṭoji's commentators says that here kṛñ is not the series. So bhū and as as helping verbs take parasmaipada: edhāmbabhūva (SK 2247), edhām āsa (SK 2248).
kṛ + ta → kṛ + e → kṛ kṛ e → ka kṛ e → ca kṛ e → cakre (edhāṃcakre)This group builds the helping verb kṛ in liṭ. SK 2241 replaces ta with e. SK 2242 makes the ending kit, so there is no guṇa. SK 2243 holds back yaṇ until the root is doubled. SK 2244–2245 reshape the first copy. SK 2246 bars iṭ, and SK 2247 turns dh into ḍh. The running example: kṛ + ta becomes kṛ + e, is doubled, and ends as cakre. Joined to edhām, this is edhāṃcakre "he grew".
Rule. In liṭ, the endings ta and jha are replaced by eś and irec, respectively. The markers are ś and c. What is heard is e and ire. The ś makes each replace the whole ending.
Example.
kṛ + ta → eś for ta → kṛ + e.
kṛ + jha → irec for jha → kṛ + ire (giving edhāṃcakrire, SK 2245).
Applies when / blocked by. SK 2233 would already make e out of a ṭi. So why does the substitute itself contain e? Bhaṭṭoji says it is a sign (jñāpaka). It shows that SK 2233 does not reach substitutes of the ātmanepada endings. So the luṭ substitutes ḍā, rau and ras (SK 2188) keep their vowels: edhitā, edhitārau, edhitāraḥ (SK 2248).
Rule. A liṭ ending that is not pit, coming after something that is not a consonant cluster (saṃyoga), is kit. "Not pit" is carried from SK 2234.
Example.
kṛ + e → kṛ ends in a vowel, not a cluster, and e is not pit → the ending is kit
→ guṇa by SK 2168 is blocked by SK 2217 → kṛ + e keeps ṛ.
Applies when / blocked by. Why "after a non-cluster"? One of Bhaṭṭoji's commentators gives mamanthe "he churned". Manth ends in a cluster, so its liṭ is not kit, and its n is not lost.
Watch out. Now kṛ + e faces a problem. Yaṇ (ṛ → r before a vowel) is due. It is later in order than doubling, so it would win. Then there would be no vowel left to double. SK 2243 solves this.
Rule. When a vowel that causes doubling follows, no substitute replaces a vowel while the doubling is still to be done. "Of a vowel" comes from SK 50, "substitute" from SK 49, "not" from SK 51. The word dvirvacane is read twice: once for "causing doubling", once for "while doubling is to be done".
Example.
kṛ + e → the e of liṭ causes doubling → yaṇ is held back
→ doubling (SK 2177) → kṛ kṛ e
→ once doubling is done, yaṇ may apply (see SK 2245, cakre).
Applies when / blocked by. Bhaṭṭoji reads the rule as a ban on the substitute. Other commentators take it as "the substitute is treated like the original"; scholars discuss this. One of Bhaṭṭoji's commentators says the result is the same either way.
Watch out. The ban lasts only until the doubling is done. After that the substitute comes as usual.
Rule. The ṛ of the abhyāsa becomes a when an affix follows. The abhyāsa is the first, repeated copy of a doubled root. The new a is followed by r (SK 70). "Of the abhyāsa" is carried from SK 2620.
Example.
kṛ kṛ e → the ṛ of the abhyāsa → a, with r after it → kar kṛ e
→ SK 2179 (halādiḥ śeṣaḥ) keeps only the first consonant of the abhyāsa → ka kṛ e.
Applies when / blocked by. Why "when an affix follows"? Bhaṭṭoji gives vavraśca "he cut", from vraśc. One of Bhaṭṭoji's commentators explains. There the abhyāsa's ṛ is itself a saṃprasāraṇa vowel. By SK 50, a vowel substitute caused by what follows it is treated like the original. Because this rule names a following affix as its cause, the new a still counts as that saṃprasāraṇa vowel. A second saṃprasāraṇa, of the v, is then forbidden, and the v stays.
Rule. A ka-series consonant or an h in the abhyāsa is replaced by a ca-series consonant. "Of the abhyāsa" is carried from SK 2620.
Example.
ka kṛ e → k of the abhyāsa → c → ca kṛ e
→ now yaṇ: ṛ → r before e → cakre
→ with the base → edhāṃcakre "he grew".
Also edhāṃcakrāte "the two grew" and edhāṃcakrire "they grew".
Applies when / blocked by. Which ca-series sound replaces which? One of Bhaṭṭoji's commentators answers. There are six sounds to replace and five substitutes. So the match goes by the outer articulation (bāhyaprayatna): voicing, breath, aspiration. Thus k → c, kh → ch, g → j, gh → jh, ṅ → ñ, and h → jh.
Rule. After a root that has one vowel and is anudātta in upadeśa, a val-initial ārdhadhātuka affix does not take iṭ. An anudātta root is one taught with the low accent. "Root" is carried from SK 2390, "not" from SK 2981. The iṭ in question is that of SK 2184.
Example.
edhām + kṛ + thās → thās becomes se (SK 2236) → doubling etc. → edhāṃ + cakṛ + se
→ se begins with s, a val consonant; iṭ would come by SK 2184
→ kṛ has one vowel and is anudātta → no iṭ
→ the s of the affix becomes ṣ → edhāṃcakṛṣe "you grew".
Also edhāṃcakrāthe "you two grew".
Applies when / blocked by. Upadeśe "in the form as taught" goes with both words. One of Bhaṭṭoji's commentators gives the reasons. If it went only with "one vowel", kartum "to do" would take iṭ, since kṛ bears the high accent in that word. If it went only with "anudātta", edhāṃcakṛṣe would take iṭ, since cakṛ has two vowels after doubling.
Why "one vowel" at all? Bhaṭṭoji says it keeps the ban out of forms of the intensive with luk (yaṅ-luk). He quotes a traditional verse. Five things do not hold in yaṅ-luk: what is cited with śtip (the ending -ti), with śap, with a marker, or by class, and any rule that names "one vowel".
Why "eka" when ac alone could mean "one vowel"? Bhaṭṭoji says the extra word excludes a root that has more than one vowel in any form in which it is taught. So vadh, the substitute of han, takes iṭ. As taught in its own right, vadh has two vowels.
Watch out. Which roots are anudātta? Bhaṭṭoji lists them shortly, at SK 2258.
Rule. After an aṅga ending in an iṇ sound, the dh of ṣīdhvam, of luṅ and of liṭ becomes retroflex, that is ḍh. The iṇ sounds are the vowels other than a, ā, and the consonants h, y, v, r, l; ṛ is one of them. The heading apadāntasya mūrdhanyaḥ (SK 210) supplies "retroflex".
Example.
edhāṃ + cakṛ + dhve (second person plural) → cakṛ ends in ṛ, an iṇ sound
→ dh → ḍh → edhāṃcakṛḍhve "you grew".
The first person: edhāṃcakre, edhāṃcakṛvahe, edhāṃcakṛmahe.
With bhū as the helping verb: edhāmbabhūva "he grew". It has the same sense as edhāṃcakre. It is parasmaipada, because SK 2240 names only kṛñ.
Applies when / blocked by. When as is the helping verb, one might expect SK 2470 to replace it with bhū before an ārdhadhātuka. Bhaṭṭoji says it does not. After-application is taught for as itself. Otherwise Pāṇini would simply have said "kṛ and as" or "kṛ and bhū". Bhaṭṭoji's commentators agree.
Watch out. The rule needs an aṅga ending in an iṇ sound. In edhiṣīdhvam (SK 2257) the dh follows the ī of sīy, but the aṅga, edh, does not end in an iṇ sound. Iṭ and sīy belong to the affix, so there is no ḍh.
edhām + as + a → edhām āsa; edh + iṭ + tās + ḍā → edhitā → edhitādhve, edhitāheThree rules finish liṭ and build the two futures of edh. SK 2248 lengthens the a of the doubled as, giving edhām āsa. Then comes luṭ, the periphrastic future. SK 2249 drops the s of tās before dh. SK 2250 turns it into h before e. Last comes lṛṭ: edhiṣyate. The running examples: edhām + as + a gives edhām āsa "he grew". edh + iṭ + tās + ḍā gives edhitā "he will grow".
Rule. The a at the beginning of an abhyāsa is lengthened. "Of the abhyāsa" is carried from SK 2620, and "long" from SK 2456. Bhaṭṭoji calls this an exception to pararūpa (SK 191).
Example.
edhām + as + liṭ → ṇal (third person singular) → edhām + as + a
→ doubling → as as a → SK 2179 keeps one consonant of the abhyāsa → a as a
→ pararūpa by SK 191 would come → instead this rule: ā as a → āsa
→ edhām āsa "he grew"; edhām āsatuḥ "the two grew".
One of Bhaṭṭoji's commentators completes the list: edhām āsuḥ, edhām āsitha, edhām āsathuḥ, edhām āsa (first person), edhām āsiva, edhām āsima.
Then Bhaṭṭoji gives luṭ, the periphrastic future:
edh + luṭ → edh + ta → tās (SK 2186) and ḍā for ta (SK 2188) → edh + tās + ā
→ iṭ (SK 2184) → edh + i + tās + ā → the ṭi ās is lost → edhitā "he will grow".
Also edhitārau, edhitāraḥ; edhitāse, edhitāsāthe.
Applies when / blocked by. Is this rule really an exception to pararūpa? Some later grammarians doubt it, pointing to another passage of Bhaṭṭoji; scholars discuss this.
Watch out. Edh takes iṭ in luṭ, although it has one vowel. SK 2258 explains: edh is not an anudātta root.
Rule. An s is lost when an affix beginning with dh follows. One of Bhaṭṭoji's commentators carries "s" from SK 2342 and "loss" from SK 2191.
Example.
edhitās + dhve (second person plural of luṭ) → dhve is an affix beginning with dh
→ the s of tās is lost → edhitādhve "you will grow".
Applies when / blocked by. Bhaṭṭoji says "when an affix follows". So in payo dhāvati "milk runs" the dh is not an affix, and nothing is lost.
Watch out. This rule returns in the aorist: in aidhiḍhvam (SK 2258) the s of sic is lost before dhvam.
Rule. Before e, the s of tās and of as "to be" becomes h. "S" is carried from SK 2342, and "of tās and as" from SK 2191. The a in ha is for pronunciation.
Example.
edh + iṭ (first person singular of luṭ) → the ṭi of iṭ becomes e → edhitās + e
→ s before e → h → edhitāhe "I shall grow".
But edhitāsvahe, edhitāsmahe: here v and m follow the s, not e, so it stays.
Then lṛṭ, the simple future:
edh + lṛṭ → edh + ta → sya (SK 2186) → edh + sya + ta
→ iṭ → edh + i + sya + ta → s becomes ṣ, the ṭi of ta becomes e → edhiṣyate "he will grow".
The paradigm: edhiṣyate, edhiṣyete, edhiṣyante; edhiṣyase, edhiṣyethe, edhiṣyadhve; edhiṣye, edhiṣyāvahe, edhiṣyāmahe.
Watch out. The rule needs e right after the s. In luṭ this happens only in the first person singular.
edhate → edhatām; edha-se → edhasva; edha-e → edhaiThe imperative, loṭ, of edh starts from the laṭ endings with their final e (SK 2233). Three rules then replace that e. SK 2251 makes it ām. SK 2252 makes it va after s and am after v. SK 2253 makes it ai in the first person. The running example: edhate becomes edhatām "let him grow". edha-se becomes edhasva "grow!". edha-e becomes edhai "let me grow".
Rule. The e of a loṭ ending is replaced by ām. "Of loṭ" is carried from SK 2198.
Example.
edh + loṭ → edh + ta → śap, and the ṭi becomes e (SK 2233) → edhate
→ this rule: e → ām → edhatām "let him grow".
In the same way: edhetām "let the two grow" (from edhete), edhantām "let them grow" (from edhante).
Watch out. The e replaced is the e that SK 2233 has just made. So loṭ shares the first steps of laṭ, and differs only at the end.
Rule. After s and after v, the e of loṭ is replaced by va and am, respectively. So after s the e becomes va; after v it becomes am. This is an exception to SK 2251.
Example.
edha + se (from thās, SK 2236) → e after s → va → edhasva "grow!".
edha + dhve (from dhvam) → e after v → am → edhadhvam "grow! (you all)".
The dual edhethām comes by SK 2251, since its e follows th.
Watch out. The substitute after s is va, the whole unit v + a, not a bare v. One of Bhaṭṭoji's commentators says so. That is why the form ends -sva.
Rule. The e of a first person (uttama) loṭ ending is replaced by ai. This is an exception to ām (SK 2251).
Example. One of Bhaṭṭoji's commentators gives the steps:
edha + e (first person singular) → e → ai → edha + ai
→ the augment āṭ (SK 2204) → edha + ā + ai
→ vṛddhi (SK 269 and vṛddhir eci) → edhai "let me grow".
Also edhāvahai, edhāmahai "let us two / us all grow".
The full loṭ of edh: edhatām, edhetām, edhantām; edhasva, edhethām, edhadhvam; edhai, edhāvahai, edhāmahai.
Watch out. The first person loṭ takes the augment āṭ (SK 2204) as well as this substitute. Both are needed for edhai.
ā + edha + ta → aidhata; edha + (s)īy + ta → edheta; edh-i-sīy-s-ta → edhiṣīṣṭa; aidh-i-ṣ-at-a → aidhiṣataThis group finishes edh. SK 2254 gives the augment āṭ to roots that begin with a vowel. SK 2255–2257 build the ātmanepada liṅ. SK 2258 gives at for jh after a non-a, which completes the aorist. Bhaṭṭoji then lists the anudātta roots, which take no iṭ. The running examples: ā + edha + ta gives aidhata "he grew". edha + īy + ta gives edheta "he should grow". edh-i-sīy-s-ta gives edhiṣīṣṭa "may he grow". aidh-i-ṣ-at-a gives aidhiṣata "they grew".
Rule. Roots that begin with a vowel take the augment āṭ in luṅ, laṅ and lṛṅ. The three lakāras are carried from SK 2206. Āṭ is heard as ā. Bhaṭṭoji calls it an exception to aṭ.
Example.
edh + laṅ → edh + ta → śap → edha + ta
→ this rule: āṭ → ā + edha + ta
→ vṛddhi by SK 269 (āṭaś ca): ā + e → ai → aidhata "he grew".
The full laṅ: aidhata, aidhetām, aidhanta; aidhathāḥ, aidhethām, aidhadhvam; aidhe, aidhāvahi, aidhāmahi.
Applies when / blocked by. Why not the ordinary aṭ (SK 2206)? One of Bhaṭṭoji's commentators answers. With aṭ, a + e would not give vṛddhi. SK 191 (ato guṇe), the later rule, would give pararūpa instead. The ā of āṭ avoids that.
Why cite SK 269, when vṛddhir eci would also give ai? The same commentator says SK 269 is needed anyway for forms like aikṣata "he saw", and is the proper rule here too.
Watch out. The laṅ endings keep their vowel: aidhata, not *\aidhate. Laṅ is not ṭ-marked, so SK 2233 does not apply. Whether this sūtra is needed at all, scholars discuss this.
Rule. The endings of liṅ receive the augment sīyuṭ. The ṭ is a marker and the u is for pronunciation, so what is heard is sīy. One of Bhaṭṭoji's commentators says it serves the ātmanepada endings. The parasmaipada endings already have yāsuṭ (SK 2209).
Example.
edh + liṅ (vidhiliṅ) → edh + ta → śap → edha + ta
→ this rule: sīy → edha + sīy + ta
→ the non-final s is lost (SK 2211) → edha + īy + ta
→ a + ī gives e (SK 69) → edhey + ta
→ y is lost before t (SK 873) → edheta "he should grow".
Also edheyātām "the two should grow", where the y stands before a vowel.
Watch out. Bhaṭṭoji's own comment is one word, salopaḥ "loss of s". In the vidhiliṅ, SK 2211 removes the s of sīy as it did the s of yās.
Rule. The jha of liṅ is replaced by ran. "Of liṅ" is carried from SK 2255. Being more than one sound, ran replaces the whole of jha.
Example.
edha + sīy + jha → ran for jha → edha + sīy + ran
→ s lost, guṇa, y lost → edheran "they should grow".
Also edhethāḥ, edheyāthām, edhedhvam (second person).
Rule. The ending iṭ of liṅ is replaced by a. "Of liṅ" is carried from SK 2255.
Example.
edha + sīy + iṭ → a for iṭ → edha + sīy + a
→ s lost, guṇa → edheya "I should grow".
Also edhevahi, edhemahi.
The full vidhiliṅ of edh: edheta, edheyātām, edheran; edhethāḥ, edheyāthām, edhedhvam; edheya, edhevahi, edhemahi.
Bhaṭṭoji then gives the benedictive (āśīrliṅ). Here the liṅ is ārdhadhātuka (SK 2215), so SK 2211 does not remove the s. The s of sīyuṭ and of suṭ belong to an affix, so both become ṣ.
edh + ta (benediction) → no śap → sīy and suṭ (SK 2210) → edh + sīy + s + ta
→ iṭ before sīy → edh + i + sīy + s + ta
→ y is lost before s → edh + i + sī + s + ta
→ both s become ṣ → edhiṣīṣta → t becomes ṭ after ṣ → edhiṣīṣṭa "may he grow".
The full benedictive: edhiṣīṣṭa, edhiṣīyāstām, edhiṣīran; edhiṣīṣṭhāḥ, edhiṣīyāsthām, edhiṣīdhvam; edhiṣīya, edhiṣīvahi, edhiṣīmahi.
Last, Bhaṭṭoji starts the aorist: aidhiṣṭa "he grew", aidhiṣātām "the two grew". Here cli becomes sic (SK 2221–2222), with iṭ, āṭ and vṛddhi.
Applies when / blocked by. Does the rule that makes ṣ apply once to both s's, or once to each? Scholars discuss this.
Watch out. edhiṣīdhvam has dh, not ḍh. Its dh follows ī, but the aṅga does not end in that sound. The i of iṭ and the ī of sīy belong to the affix. So SK 2247 does not apply.
Rule. In an ātmanepada ending, a jh that comes after something other than a is replaced by at. "Jh" is carried from SK 2169, and "at" from SK 2479. This rule comes where SK 2169 would give ant.
Example.
edh + luṅ → edh + jha → sic (SK 2221–2222), iṭ, āṭ, vṛddhi → aidh + i + s + jha
→ the jh follows s, not a → at for jh → aidh + i + s + ata
→ s becomes ṣ → aidhiṣata "they grew".
Also aidhiṣṭhāḥ "you grew", aidhiṣāthām "you two grew".
The second person plural:
aidh + i + s + dhvam → s lost before dh (SK 2249) → aidhidhvam
→ dh → ḍh after an iṇ-final aṅga (SK 2247) → aidhiḍhvam.
The first person: aidhiṣi, aidhiṣvahi, aidhiṣmahi.
The full aorist of edh: aidhiṣṭa, aidhiṣātām, aidhiṣata; aidhiṣṭhāḥ, aidhiṣāthām, aidhiḍhvam; aidhiṣi, aidhiṣvahi, aidhiṣmahi.
The conditional, lṛṅ: aidhiṣyata, aidhiṣyetām, aidhiṣyanta; aidhiṣyathāḥ, aidhiṣyethām, aidhiṣyadhvam; aidhiṣye, aidhiṣyāvahi, aidhiṣyāmahi.
Applies when / blocked by. Why "in ātmanepada"? One of Bhaṭṭoji's commentators gives adanti "they eat" and sunvanti "they press out": parasmaipada, so ant. Why "after a non-a"? edhante, plavante "they float": there jh follows a, so ant.
In aidhiḍhvam, one of Bhaṭṭoji's commentators says the iṭ counts as part of the aṅga ending in sic. So the aṅga ends in an iṇ sound, and SK 2247 applies. Others hold that the iṇ of SK 2247 does not include iṭ. On their view the form is aidhidhvam. Bhaṭṭoji records this view with a "but" (tu). One of Bhaṭṭoji's commentators reads that as a sign that the view lacks the Bhāṣya's support. The Bhāṣya is Patañjali's great commentary on Pāṇini's sūtras.
Bhaṭṭoji adds that ḍh or dh, v and m may each be doubled or not. So there are sixteen forms of this one word. One of Bhaṭṭoji's commentators shows that the count includes both views: ḍh and dh.
Why does edh take iṭ at all? Bhaṭṭoji says: because edh is udātta. One of Bhaṭṭoji's commentators meets an objection. Edh is taught with an anudātta marker vowel. But once that marker is gone, the root that remains is not anudātta. So SK 2246 does not bar iṭ.
Watch out. At is for jh after a non-a. After a, as in the present edhante, the jh becomes ant as usual.
After this rule Bhaṭṭoji collects the anudātta roots, "incidentally", in eleven verses. These are the roots of SK 2246, which take no iṭ before a val-initial ārdhadhātuka.
Verse 1, the vowel-final roots. Every root of one vowel that ends in a vowel is anudātta. The exceptions are the roots in ū and ṝ, and yu, ru, kṣṇu, śīṅ, snu, nu, kṣu, śvi, ḍīṅ, śri, vṛṅ and vṛñ. So, one of Bhaṭṭoji's commentators notes, dātā, dhātā, cetā, stotā have no iṭ. But bhū ends in ū, so it takes iṭ: bhavitā.
Verses 2–6, the consonant-final roots. There are a hundred and two, cited as Bhaṭṭoji cites them:
Verses 7–9, why some are cited oddly. Some roots have look-alikes in other classes that are not anudātta. To exclude those, Bhaṭṭoji cites the right root with its vikaraṇa or its markers. Tṛp and dṛp stand also in the tudādi and curādi classes (the root classes headed by tud and by cur), on some views. So they are cited with śyan: tṛpyati, dṛpyatī. Likewise svid, pad, sidh, budh, man, puṣ and śliṣ are cited with śyan. Vas is cited with śap (vasati), and yu with luk (yauti). Ṇijir, vijir and śaklṛ are cited with their markers.
Verses 10–11, two notes on authority. Vindati "finds" is accepted as a root without iṭ by other grammars, and is seen in the Bhāṣya. Vyāghrabhūti and others, earlier authors of such verses, did not list it. One of Bhaṭṭoji's commentators concludes that it takes no iṭ, on the Bhāṣya's authority. Nine of the listed roots are not named in the Bhāṣya: rañj, masj, ad, pad, tud, kṣudh, śuṣ, puṣ, śiṣ. They are included on the authority of Vyāghrabhūti and others.
Last, Bhaṭṭoji gives root 3 of the list, spardha "to rival" (saṃgharṣe). Rivalry is the wish to overcome another. Its object is contained in the root's own meaning, so the root is akarmaka, "without an object"; scholars discuss this. Like edh, it is an anudāttet root and takes ātmanepada: spardhate "he rivals".
spardh + e → paspardhe; dadh dadh e → dedhe; skud → skundate, cuskunde; dad → dadadeThe edh paradigm is finished, and Bhaṭṭoji now walks down the list of roots (the dhātupāṭha) that follow edh. They take ātmanepada and end in d or dh. Each new root shows one new rule of liṭ, the perfect. In spardh "to compete", the reduplicated syllable keeps p, not s: paspardhe. In dadh "to hold", the root's a becomes e and the reduplicated syllable disappears: dedhe. In skud "to leap up", an n is inserted: skundate. And dad "to give" refuses the e-rule: dadade, not *dede.
Rule. In the abhyāsa, the first, reduplicated syllable of a doubled root, a khay that stands after a śar is the consonant that remains. A khay is a hard stop such as p, t, k or c. A śar is a sibilant: ś, ṣ or s. This rule is an exception (apavāda) to SK 2179, halādiḥ śeṣaḥ, which keeps only the first consonant of the abhyāsa.
Example.
spardh + e (liṭ, prathama-puruṣa singular ātmanepada)
→ the root is doubled: spardh spardh e
→ SK 2179 alone would keep only s in the abhyāsa and give *saspardhe
→ SK 2259: the khay p, which follows the śar s, remains instead
→ pa spardh e → paspardhe.
The other lakāras of spardh are formed like those of edh. The only difference is that a consonant-initial root takes the augment aṭ, not āṭ. So: spardhitā (luṭ), spardhiṣyate (lṛṭ), spardhatām (loṭ), aspardhata (laṅ), spardheta (vidhiliṅ), spardhiṣīṣṭa (āśīrliṅ), aspardhiṣṭa (luṅ), aspardhiṣyata (lṛṅ).
Applies when / blocked by. Bhaṭṭoji's commentators read śarpūrvāḥ "having a śar before them" as a special kind of bahuvrīhi compound. In it the thing named in the compound, the śar, is not itself included (an atadguṇasaṃvijñāna bahuvrīhi). So the śar goes and only the khay remains. The word ādiḥ "initial" is also carried forward from SK 2179. The khay must be the first consonant once the śar is set aside. That is why the root vraśc keeps no c in its abhyāsa.
Roots listed after this rule. Root 4 gādhṛ "to stand firm": gādhate, liṭ jagādhe. Root 5 bādhṛ "to oppress": bādhate. Roots 6 nāthṛ and 7 nādhṛ "to request". A vārttika (a supplementary statement) says: āśiṣi nāthaḥ, nāth takes ātmanepada in the sense of blessing. The root already takes ātmanepada through its it-vowel. So Bhaṭṭoji's commentators read the vārttika as a restriction (niyama): ātmanepada only in the sense of blessing. So nāthate "he blesses", but elsewhere nāthati. Then nādhate, and root 8 dadha "to hold": dadhate.
Watch out. The kept consonant is the khay, never the sibilant. Do not write *saspardhe or *spaspardhe.
Rule. Before a kit liṭ ending, an a that stands between two single consonants becomes e. At the same time the abhyāsa is deleted. A kit affix is one marked with k, which blocks guṇa; by SK 2242, a liṭ ending that is not pit is kit after a root that does not end in a conjunct. The words "e" and "deletion of the abhyāsa" are carried forward from SK 2471.
There is one more condition. The aṅga must not begin with a substitute that liṭ itself has brought about. The word liṭi is read twice: once for that substitute, once for the following liṭ.
Example. Bhaṭṭoji gives the examples under the next rule. The pattern is:
dadh + e (kit liṭ)
→ dadh dadh e
→ the a of the root stands between single consonants, d and dh
→ it becomes e, and the abhyāsa is deleted
→ dedhe.
Applies when / blocked by. The a must be between single consonants. The word eka here means "unaccompanied", that is, not part of a conjunct. One of Bhaṭṭoji's commentators gives two counter-examples. Why "kit"? nanāda, where the ending ṇal is not kit. Why "liṭ"? pāpacyāt, which is not a liṭ form.
Watch out. The rule does two things at once. The vowel changes to e and the whole abhyāsa disappears. dedhe has no reduplicated syllable left.
Rule. The same two changes, e for a and loss of the abhyāsa, also come before the ending thal when it has the augment iṭ. The ending thal (-tha, second person singular parasmaipada) is not kit. So SK 2260 would not reach it; this rule extends it.
Example.
dadh + e → dadh dadh e → dedhe; likewise dual dedhāte, plural dedhire.
For thal with iṭ, a root whose n is not a liṭ substitute (ṇam, see below):
nam + i + tha → e for a, no abhyāsa → nemitha.
Applies when / blocked by. Here Bhaṭṭoji explains the word ādeśa "substitute" in SK 2260. It means only a substitute that changes the sound. The proof (jñāpaka) is SK 2263, which names śas and dad. If every substitute blocked the rule, the abhyāsa of those roots would already block it, and SK 2263 would be pointless. So in dadh dadh e the abhyāsa's d, even if it counts as a substitute, is the same sound. The rule applies: dedhe.
Bhaṭṭoji lists what each word of SK 2260 excludes:
didivatuḥ (from div): no a stands there.rarāse (from rās): the vowel is long.tatsaratuḥ (from tsar): the a follows a conjunct.cakaṇatuḥ (from kaṇ): the abhyāsa's k has become c, a different sound.Since the substitute must be caused by liṭ, the rule does apply to nemitha and sehe. The n of nam comes from SK 2286, and the s of sah from SK 2264. Neither is caused by liṭ.
Roots listed after this rule. Root 9 skudi "to leap up, to lift out" leads into the next rule.
Rule. A root whose it-marker is short i takes the augment num. Such a root is called idit. In num, the m is an it-marker and the u is only for pronunciation. What is added is n. Being marked with m, it goes after the root's last vowel.
Example.
skudi → the final i is an it-marker, so the root is skud, and it is idit
→ num after the last vowel: sku-n-d
→ with śap and the ending: skundate.
liṭ: skund skund e → the abhyāsa keeps k by SK 2259 → k becomes c by SK 2245 → cuskunde.
Applies when / blocked by. Bhaṭṭoji's commentators say the i must be final in the cited root. A root like cakṣiṅ, whose i is not final, is not idit.
Roots listed after this rule. Every one of these is idit and so takes num:
klindate caitram "he grieves for Caitra"; ciklinde.Then two roots without the i: 16 muda "to rejoice": modate. 17 dada "to give": dadate.
Watch out. num changes the shape of the root for later rules. In mamande the a now stands before a conjunct, nd. So SK 2260 cannot give e.
Rule. The e and the loss of the abhyāsa taught in SK 2260 do not come for four kinds of a. They are the a of śas "to injure" and of dad "to give". Then the a of any root beginning with v. Last, an a that the word guṇa has produced.
Example.
dad + e → dad dad e
→ SK 2260 would give *dede
→ SK 2263 forbids it
→ dadade; dual dadadāte, plural dadadire.
Applies when / blocked by. The last item needs care. Guṇa of ṛ gives ar. Guṇa of u gives o, which can then become av. Bhaṭṭoji says the a meant is the one "brought about by the word guṇa". Bhaṭṭoji's commentators give paparatuḥ from pṝ: its a is part of the ar made by guṇa. One of them adds śaśasatuḥ (śas) and vavamatuḥ (a v-initial root).
Roots listed after this rule. 18 ṣvada and 19 svarda "to taste". Each is transitive in the sense "to experience" and intransitive in the sense "to be pleasing".
Watch out. The v of "v-initial" is v itself. A root that begins with b, such as bad (SK 2284), is not covered and does get e: bedatuḥ.
ṣvad → svadate, sasvade; urd → ūrdate, ūrdāṃcakreMany roots are taught in the dhātupāṭha with an initial ṣ, which is never heard as such. SK 2264 turns it into s. The ṣ is kept in the list for a reason: so that it can come back in forms like asiṣvadat, where the s is a substitute. The second rule lengthens an i or u before a root's penultimate r: urd "to measure" gives ūrdate. The group closes the long list of ātmanepada roots and opens the parasmaipada roots with at "to go constantly".
Rule. A ṣ at the beginning of a root, as the root is taught (in upadeśa), is replaced by s. The word upadeśe is carried forward from SK 2370.
Example.
ṣvad (root 18) → initial ṣ becomes s → svad → svadate.
liṭ: svad svad e → sasvade.
With the prefix anu: anu + svadate → anusvadate.
Applies when / blocked by. In anusvadate the new s follows the u of anu. A substitute s after such a vowel would normally turn back into ṣ. Bhaṭṭoji says SK 2123, sāt padādyoḥ, forbids that here, since the s begins a pada.
In sasvade the a stands after a conjunct, sv. So there is no e by SK 2260, Bhaṭṭoji's commentators note. Likewise svardate, liṭ sasvarde.
Why "at the beginning"? laṣati: the ṣ is not initial and stays.
Watch out. Why teach the root with ṣ at all, if it always becomes s? One of Bhaṭṭoji's commentators answers. Only an s that is a substitute can later turn into ṣ. So the ṣ is heard again in a form like asiṣvadat. A root taught with plain s would never show it.
Rule. An ik vowel (i, u, ṛ or ḷ) is lengthened before r or v. That r or v must be the root's penultimate (upadhā), followed by a consonant. The words "r or v", "ik" and "long" come from SK 433; "of a root" from SK 2485; "before a consonant" from SK 354.
Example.
urd + a + te (root 20 urda "to measure, to play")
→ the u stands before r; r is the root's penultimate, followed by d
→ u becomes ū → ūrdate.
liṭ with ām (SK 2237): ūrdāṃcakre.
Applies when / blocked by. Bhaṭṭoji's commentators give three counter-examples. No r or v: puṣpyati. The r or v not followed by a consonant: ciriṇoti. The vowel not an ik: nardati.
Roots listed after this rule. 21 kurda, 22 khurda, 23 gurda, 24 guda "to play": kūrdate, liṭ cukūrde, gūrdate. guda has no r, so no lengthening: godate, jugude. 25 ṣūda "to flow": sūdate, liṭ suṣūde.
Here Bhaṭṭoji quotes a verse that tells you which s-initial roots were taught with ṣ (ṣopadeśa). A monosyllabic root beginning with s, followed at once by a purely dental sound or by a vowel, is ṣopadeśa. Seven are excepted: sek, sṛp, sṛ, stṛ, sṛj, stṝ, styā. Six are added: ṣvaṣk, svid, svad, svañj, svap, smiṅ. "Dental" means purely dental, not the teeth-and-lips sound v. That is shown by the separate mention of ṣvaṣk and the rest.
More roots follow: hrāda (jahrāde), hlādī, svāda, parda. Then yatī "to strive": yatate, liṭ yete, with e and loss of the abhyāsa by SK 2260. Then yutṛ (yuyute), jutṛ (jujute), vithṛ (vivithe), vethṛ (vivethe), śranthate, granthate, katthate. Bhaṭṭoji then says: the ātmanepada roots beginning with edh are done.
Next come thirty-eight parasmaipada roots ending in a t-class consonant. The first is 38 ata "to go constantly": atati. In liṭ, SK 2248 lengthens the abhyāsa's a: āta, ātatuḥ, ātuḥ. In luṅ the form reaches the stage āt-i-s-ī-t. The next group finishes it.
Watch out. svāda "to taste" is not ṣopadeśa. Its s is followed by v, which is not purely dental.
āt-i-s-ī-t → ātīt; mā bhavān atīt; cyut → acyutat / acyotītA root that takes the augment iṭ is called seṭ; one that does not is aniṭ. The aorist (luṅ) of a seṭ root is built as a + root + i + s + ī + t. Here s is the affix sic, i is the augment iṭ, and ī is the augment īṭ. This group shows three things about such forms. The s drops between iṭ and īṭ, so āt-i-s-ī-t becomes ātīt. The vṛddhi that consonant-final roots take before sic is blocked when iṭ follows. And a root marked with ir may form its aorist with the affix aṅ instead: acyutat beside acyotīt.
Rule. An s that comes after iṭ is elided when īṭ follows. "s" is carried forward from SK 280, and "elision" (lopa) from SK 54.
Example.
āt-i-s-ī-t (luṅ of at, from SK 2265)
→ the s stands after iṭ (i) and before īṭ (ī): it is elided
→ āti-ī-t
→ i and ī merge into ī → ātīt.
Applies when / blocked by. There is a difficulty. The loss of the s counts as not yet done (asiddha) for the merging of the vowels. So i and ī could not merge. A vārttika solves this: the loss of sic is to be treated as done (siddha) for a single replacement (ekādeśa). So the two vowels can merge.
In ātiṣṭām and ātiṣuḥ no īṭ follows. So the s stays, and becomes ṣ. One of Bhaṭṭoji's commentators gives the other counter-example: why "after iṭ"? ahārṣīt, where no iṭ precedes.
Watch out. The rule needs both augments, iṭ before and īṭ after. Only two forms of the paradigm, ātīt and ātīḥ, have īṭ.
Rule. The vowel of vad, of vraj, and of any aṅga ending in a consonant takes vṛddhi. This happens before sic followed by a parasmaipada ending. The words "sic", "vṛddhi" and "parasmaipada" are carried forward from SK 2297, taught later in this chapter.
Example. Bhaṭṭoji cites the rule here to raise a problem. By it, the a of at in āt-i-s-ī-t would take vṛddhi. The next rule answers.
Consonant-final roots with vṛddhi (one of Bhaṭṭoji's commentators): apākṣīt (pac), adhākṣīt (dah).
Applies when / blocked by. vad and vraj end in consonants anyway. Why name them? Bhaṭṭoji's commentators answer: so that their vṛddhi is obligatory. Otherwise the option of SK 2284 would reach them. So avādīt and avrājīt only.
Rule. The vṛddhi just taught does not come when the sic begins with iṭ. Bhaṭṭoji's commentators add that this blocks only the vṛddhi of consonant-final roots. The vṛddhi named for vad and vraj still comes.
Example.
mā + at + luṅ (with the particle mā, which blocks āṭ)
→ at-i-s-ī-t → no vṛddhi, since the sic begins with iṭ
→ sic lost by SK 2266 → mā bhavān atīt "may your honour not go".
Dual atiṣṭām, plural atiṣuḥ.
Applies when / blocked by. In ātīt the augment āṭ merges with the root's a anyway. So vṛddhi or no vṛddhi makes no difference there. The effect of this rule shows only with mā, where there is no āṭ, one of Bhaṭṭoji's commentators says.
The consonant-final vṛddhi still comes when sic has no iṭ: adhākṣīt.
Roots listed after this rule. 39 citī "to perceive": cetati, liṭ ciceta, luṅ acetīt, acetiṣṭām, acetiṣuḥ (no vṛddhi, by this rule). 40 cyutir "to sprinkle". A vārttika says that ir as a whole is the it-marker. So cyutir is not idit and takes no num: cyotati, liṭ cucyota.
Watch out. Do not read cyutir as idit because it contains an i. The ir is one it-marker.
Rule. After a root marked with ir (an irit root), the affix aṅ optionally replaces cli, the general aorist affix, before a parasmaipada ending. Otherwise cli becomes sic as usual. "cli" comes from SK 2222, "aṅ" from SK 2438, "parasmaipada" from SK 2343.
Example.
a + cyut + cli + t
→ with aṅ: a + cyut + a + t → aṅ is marked with ṅ, so no guṇa → acyutat
→ without aṅ: a + cyot + i + s + ī + t → sic lost by SK 2266 → acyotīt.
Roots listed after this rule.
aścyutat / aścyotīt. It is also found without y: ścotati.mathyāt.kunthyāt, manthyāt.asedhīt.After a prefix, the s of sedh would stay s by SK 2123, sāt padādyoḥ. That is why the next group of rules is needed.
Watch out. The option is only for parasmaipada endings. aṅ carries a ṅ, so it never allows guṇa: acyutat has u, acyotīt has o.
ni + sedhati → niṣedhati; ni-a-sedhat → nyaṣedhat; ni + sisedha → niṣiṣedha; but gaṅgāṃ visedhatiWhen a prefix (upasarga) such as ni or vi stands before a root beginning with s, the s often becomes ṣ. The general rule for this cannot apply, because the s begins a new pada (SK 2123). So Pāṇini names the roots one by one. This group gives the lists, then two extensions. The ṣ comes even when the augment a (aṭ) stands between prefix and root: nyaṣedhat. For ten roots it also comes across the reduplicated syllable: niṣiṣedha. Last comes one exception: sedh meaning "to go" keeps its s, as in gaṅgāṃ visedhati "he goes to the Gaṅgā".
Rule. After a cause standing in an upasarga, s becomes ṣ in eleven roots: sunoti, suvati, syati, stauti, stobhati, sthā, senaya, sedha, sic, sañj and svañj. The cause is an iṇ sound, such as the i of ni or abhi. "s" is carried forward from SK 335, and the ṣ comes under the heading SK 210.
Example.
abhi + sunoti → the i of abhi is the cause → s becomes ṣ (n → ṇ by SK 197) → abhiṣuṇoti.
ni + sedhati (root 47 ṣidha) → niṣedhati.
Applies when / blocked by. Bhaṭṭoji glosses "upasarga" as "standing in an upasarga". The cause is the prefix's sound, not the prefix as a word.
Several roots are named in a special form. sedha, with the vowel of śap, means the bhvādi root, not the divādi sidhyati. senaya is a nominal root (from senā) ending in the affix ṇic. The others are named with the ending -ti. The older authorities say this excludes the intensive forms with luk of yaṅ; scholars discuss this.
Watch out. This rule is needed only because SK 2123 would block the ṣ. The s of sedhati begins its own pada after the prefix.
Rule. The s of sad "to sit" becomes ṣ after any upasarga except prati.
Example. Bhaṭṭoji gives none here. One of Bhaṭṭoji's commentators gives the counter-example: pratisīdati, with s, because the prefix is prati. With ni you will meet niṣīdati later in this chapter.
Applies when / blocked by. Bhaṭṭoji explains these ṣ-rules in their own order. The order matters, because SK 2276 counts back through the roots named here.
Rule. The s of the root stanbh becomes ṣ after an upasarga. Bhaṭṭoji writes the root as stanbh, with n, and calls it sautra. A sautra root is one taught in a sūtra of the grammar, not in the dhātupāṭha.
Example.
prati + stambha → the i of prati → pratiṣtambha → t becomes ṭ after ṣ → pratiṣṭambha, as in bāhupratiṣṭambhavivṛddhamanyuḥ "whose anger swelled at the checking of his arm".
Applies when / blocked by. Why is this a separate rule and not part of SK 2271? Bhaṭṭoji gives two reasons. First, the split is for the next rule: only stanbh, not sad, continues into SK 2273. Second, "except prati" does not continue into this rule. The example above, with prati, shows it.
The root meant is the sautra stanbh, not the dhātupāṭha root ṣṭabhi. So vistambhate, from ṣṭabhi, takes no ṣ. Bhaṭṭoji's commentators explain: a form directly taught is preferred to one produced by rule.
Watch out. The printed text of the rule reads stanbheḥ, with n; that n is what points to the sautra root.
Rule. After ava, the s of stanbh becomes ṣ in two senses: support (ālambana) and nearness (āvidūrya).
Example. Bhaṭṭoji gives none. Bhaṭṭoji's commentators give:
ava + stabhya "support" → yaṣṭim avaṣṭabhya tiṣṭhati "he stands leaning on a staff".
ava + stabdhā "nearness" → avaṣṭabdhā gauḥ "the cow kept close by".
Applies when / blocked by. The a of ava is not an iṇ sound. So no ṣ would come by SK 2272, which needs an iṇ in the upasarga; this rule is a fresh injunction. In other senses there is no ṣ. One of Bhaṭṭoji's commentators gives avastabdho vṛṣalaḥ śītena.
Rule. After vi, and by the word ca also after ava, the s of svan becomes ṣ in the sense "eating".
Example. From Bhaṭṭoji's commentators:
vi + svanati "eats noisily" → viṣvaṇati (the n becomes ṇ by SK 197).
ava + svanati → avaṣvaṇati.
Applies when / blocked by. When the sense is not eating, the s stays: visvanati vīṇā "the vīṇā sounds".
Rule. After pari, ni and vi, s becomes ṣ in eight items. They are the root sev "to serve", the participle sita "bound", the noun saya, the roots sivu and saha, the augment suṭ, and the roots stu and svañj.
Example. Bhaṭṭoji's example is niṣedhati. There the ṣ in fact comes by SK 2270, which names sedha. Bhaṭṭoji's commentators give examples for this rule itself: pariṣevate, viṣitaḥ, viṣayaḥ, pariṣīvyati, pariṣahate, pariṣkaroti (with the augment suṭ).
Applies when / blocked by. stu and svañj were already named in SK 2270. Bhaṭṭoji's commentators say they are named again for an option taught later (SK 2359).
Watch out. Bhaṭṭoji quotes this rule here because the next rule counts "up to sita", the second word of its list.
Rule. For the roots named before the word sita, ṣ comes even when the augment aṭ stands between the prefix and the s. Bhaṭṭoji's commentators count them: fifteen. They are the eleven of SK 2270, then sad, stanbh, svan and sev.
Example.
ni + sedhat (laṅ) → aṭ is added before the root: ni + a + sedhat
→ the a separates the i of ni from the s
→ this rule still gives ṣ → ni a ṣedhat
→ i becomes y before a → nyaṣedhat.
Likewise luṅ nyaṣedhīt, lṛṅ nyaṣedhiṣyat, and abhyaṣuṇot.
Applies when / blocked by. The word api "even" matters. It shows that the ṣ of SK 2270 is not limited to forms with aṭ.
Watch out. Without this rule the s would not follow the iṇ sound directly. Then no ṣ would come.
Rule. The rule has two parts. First, for the roots from sthā up to (not including) sita, ṣ comes even when the abhyāsa stands between prefix and root. Second, the s of the abhyāsa itself becomes ṣ only for these roots. Bhaṭṭoji's commentators count ten roots: sthā, senaya, sedha, sic, sañj, svañj, sad, stanbh, svan and sev.
Example.
ni + sedh + a (liṭ) → doubling: ni + si sedh a
→ the abhyāsa si now stands between ni and the s of sedh
→ part one: the root's s still becomes ṣ
→ the abhyāsa's s becomes ṣ after the i of ni (SK 2270), allowed by part two
→ niṣiṣedha; dual niṣiṣidhatuḥ.
Applies when / blocked by. Bhaṭṭoji's commentators explain part two. The abhyāsa's s already gets ṣ from SK 2270. So "of the abhyāsa" is a restriction (niyama): the abhyāsa's s becomes ṣ only for these ten. For other roots it does not: abhisusūṣati.
Watch out. The first five roots of SK 2270 (sunoti to stobhati) are not among the ten. They get ṣ across aṭ, but their abhyāsa keeps s.
Rule. The root sedh, when it means "to go", does not take ṣ. The negative "na" is carried forward from SK 3168.
Example.
vi + sedhati "goes" → no ṣ → gaṅgāṃ visedhati "he goes to the Gaṅgā".
Applies when / blocked by. Which ṣ does this rule cancel? One of Bhaṭṭoji's commentators says only the ṣ of SK 2270. He appeals to the maxim anantarasya: a prohibition applies to the nearest rule. So siṣedha still has ṣ, by SK 212. Others hold that every ṣ is cancelled; scholars discuss this.
Roots listed after this rule. Root 48 ṣidhū "to command, to be auspicious". It inflects like ṣidha, but its ū-marker brings the next group.
sisedh + tha → siṣedh-dha → siṣeddha / siṣedhitha; asaidh-s-tām → asaiddhāmṣidhū is marked with ū. Such a root takes the augment iṭ only optionally. So it has two sets of forms: siṣedhitha with iṭ, siṣeddha without. The forms without iṭ bring dh and t together. SK 2280 turns the t or th into dh, and SK 2281 drops an s squeezed between two consonants.
Rule. The augment iṭ is optional before an ārdhadhātuka affix that begins with a val consonant (such as t, th, s, v or m). This holds after svṛ, after sū (both sūti and sūyati), after dhūñ, and after any root marked with ū (an ūdit root). The rule qualifies SK 2184, which gives iṭ to such affixes.
Example. Bhaṭṭoji gives the forms under the next rule. ṣidhū is ūdit, so:
siṣedh + thal → with iṭ: siṣedhitha; without iṭ: see SK 2280.
Applies when / blocked by. Bhaṭṭoji's commentators identify the roots. svarati is svṛ of the bhvādi class. sūti and sūyati are sū of the adādi and divādi classes, not the tudādi root. dhūñ, with its ñ, excludes another root dhū.
Rule. A t or th that comes after a jhaṣ consonant, a voiced aspirate such as dh or bh, becomes dh. The root dhā is excepted.
Example.
siṣedh + tha (liṭ, madhyama-puruṣa singular, without iṭ)
→ th after dh → dh: siṣedh + dha
→ the first dh becomes d (jaśtva, SK 52) → siṣeddha.
With iṭ: siṣedhitha.
luṭ: seddhā / sedhitā. lṛṭ: setsyati / sedhiṣyati. luṅ: asaitsīt.
Applies when / blocked by. Why "except dhā"? One of Bhaṭṭoji's commentators gives dhattaḥ.
In asaitsīt there is no iṭ. īṭ comes by SK 2225 and vṛddhi by SK 2267. Then dh becomes t before s (cartva), Bhaṭṭoji's commentators explain.
Watch out. jaśtva replaces a consonant by the unaspirated voiced stop of its class. cartva replaces it by the unaspirated voiceless one. You need both to get siṣeddha and asaitsīt.
Rule. An s that comes after a jhal consonant is elided when another jhal follows. A jhal is a stop, a sibilant or h, as against a nasal or a semivowel. "s" and "elision" are carried forward as in SK 2266. One of Bhaṭṭoji's commentators adds that all three must belong to one pada.
Example.
a + saidh + s + tām (luṅ dual, without iṭ)
→ s stands between dh and t, both jhal → s elided
→ asaidh + tām → t becomes dh (SK 2280), dh becomes d → asaiddhām.
The full paradigm without iṭ: asaitsīt, asaiddhām, asaitsuḥ; asaitsīḥ, asaiddham, asaiddha; asaitsam, asaitsva, asaitsma. The s drops only before t or th.
With iṭ: asedhīt, asedhiṣṭām. There SK 2268 blocks vṛddhi.
Applies when / blocked by. SK 2266 cannot drop this s. There is no iṭ, no īṭ and no single-sound ending. One of Bhaṭṭoji's commentators gives the counter-examples. Why "after a jhal"? anaiṣṭām. Why "before a jhal"? asaitsīt.
Roots listed after this rule. 49 khādṛ "to eat" (the ṛ is an it-marker): khādati, cakhāda. 50 khada "to be firm, to injure, to eat": khadati.
Watch out. The s of sic stays before a vowel or a nasal. asaitsuḥ and asaitsma keep it.
ca-khad + ṇal(a) → cakhāda; uttama cakhāda / cakhada; akhādīt / akhadītRoot 50, khad, has a short a before its final consonant. Three rules decide when that a becomes ā. Before the liṭ ending ṇal it takes vṛddhi: cakhāda. In the first person (uttama) singular this is optional: cakhāda / cakhada. In the aorist with iṭ it is optional too: akhādīt / akhadīt.
Rule. A short a that is the penultimate (upadhā) of the aṅga takes vṛddhi before an affix marked with ñ or ṇ. "vṛddhi" is carried forward from SK 2473.
Example.
khad + ṇal → doubling: ca khad a
→ ṇal is marked with ṇ; the penultimate a becomes ā
→ cakhāda.
Applies when / blocked by. SK 254 gives vṛddhi only to a final vowel. khad ends in a consonant, so this rule is needed.
Why no e with loss of the abhyāsa (SK 2260)? Bhaṭṭoji's commentators give reasons. ṇal is marked with p, so it is not kit. The abhyāsa ca begins with a changed sound. And vṛddhi is the later rule.
One of Bhaṭṭoji's commentators gives counter-examples. Why "a"? tutoda. Why "penultimate"? gaṇayati.
Rule. The ending ṇal of the first person (uttama) is only optionally treated as marked with ṇ. "marked with ṇ" is carried forward from SK 284.
Example.
ca khad + a (uttama singular)
→ as ṇit: vṛddhi by SK 2282 → cakhāda
→ not as ṇit: no vṛddhi → cakhada.
Watch out. The prathama singular has only cakhāda. The option belongs to the uttama singular.
Rule. A light short a in an aṅga that begins with a consonant optionally takes vṛddhi. This happens before a sic that begins with iṭ and is followed by a parasmaipada ending. The words come from SK 2297 ("sic", "vṛddhi", "parasmaipada"), SK 2268 ("iṭ") and SK 2449 ("optionally").
Example.
a + khad + i + s + ī + t
→ with vṛddhi: akhādīt; without: akhadīt (sic lost by SK 2266).
Applies when / blocked by. One of Bhaṭṭoji's commentators gives counter-examples. Not a: adevīt. Not light: arakṣīt. sic without iṭ: apākṣīt.
Roots listed after this rule. 51 bada "to be steady". It begins with the labial b, not with v.
babāda (vṛddhi by SK 2282).bedatuḥ, beditha: e with loss of the abhyāsa (SK 2260, 2261). The abhyāsa's b is the same sound, so it does not block the rule.babāda / babada (SK 2283).abādīt / abadīt (this rule).52 gada "to speak clearly": gadati.
Watch out. If bada began with v, SK 2263 would block the e. Then bedatuḥ could not be formed. That is why Bhaṭṭoji says it begins with a labial.
pra + ni + gadati → praṇigadati; ṇad → nad-Two rules about n and ṇ close the part. After pra, the n of the prefix ni becomes ṇ before certain roots: praṇigadati. And a root taught with initial ṇ is always pronounced with n. Yet it remembers its ṇ: after pra, the ṇ returns.
Rule. The n of the prefix ni becomes ṇ after an r or ṣ that stands in an upasarga. This is obligatory before these roots: gada, nada, pata, pada, the ghu roots (dā and dhā), mā, syati, hanti, yāti, vāti, drāti, psāti, vapati, vahati, śāmyati, cinoti and degdhi. "n becomes ṇ after r or ṣ" is carried forward from SK 235, "upasarga" from SK 2287.
Example.
pra + ni + gadati → the r of pra is the cause → the n of ni becomes ṇ → praṇigadati.
liṭ of gad: jagāda.
Applies when / blocked by. pra and ni are separate words. So ṇ would not otherwise come, Bhaṭṭoji's commentators say. Compare SK 2232, where the ṇ for ni is only optional.
The roots are named with śap or with the ending -ti. This excludes the intensive forms with luk of yaṅ, Bhaṭṭoji's commentators say. For the ghu roots one of them gives praṇidadāti and praṇidadhāti.
Roots listed after this rule. 53 rada "to scratch": liṭ rarāda, redatuḥ. 54 ṇada "to make an inarticulate sound" leads into the next rule.
Rule. A ṇ at the beginning of a root becomes n. "At the beginning of a root" is carried forward from SK 2264.
Example.
ṇad (root 54) → initial ṇ becomes n → nad → nadati.
Applies when / blocked by. The ṇ must be root-initial. So aṇati keeps its ṇ, Bhaṭṭoji's commentators say.
Which n-initial roots were taught with ṇ (ṇopadeśa)? Bhaṭṭoji gives the answer that follows from the Mahābhāṣya. All n-initial roots are ṇopadeśa except eight: nard, nāṭi, nāth, nādh, nand, nakk, nṝ and nṛt. The excluded nāṭi is the curādi root, with long ā. So the ghaṭādi root naṭ "to dance" is ṇopadeśa after all. Some teachers exclude only five, counting nādh, nṝ and nand as ṇopadeśa too. Scholars discuss this.
Watch out. A ṇopadeśa root looks exactly like any other n-initial root. You can tell it only from this list. It matters in the next rule.
ard + ṇal → ā ard a → ānarda; ukh → uvokha, ūkhatuḥThe list of roots goes on, from root 55 to about root 196. Four rules come up on the way. A ṇopadeśa root takes ṇ after pra: praṇadati. A root beginning with a, with two consonants, gets an extra n in liṭ: ānarda. Initial ñi, ṭu and ḍu of a root are it-markers. And the i or u of an abhyāsa becomes iy or uv before an unlike vowel: uvokha.
Rule. The n of a ṇopadeśa root becomes ṇ after an r or ṣ standing in an upasarga. This holds whether or not the prefix and root form a compound. The heading is SK 235, raṣābhyāṃ no ṇaḥ.
Example.
pra + nadati (nad, taught as ṇad)
→ the r of pra is the cause; the n of nad is the target
→ praṇadati.
Applies when / blocked by. pra and nadati are separate padas. So only this rule gives ṇ. Bhaṭṭoji's commentators explain the wording. The word samāse "in a compound" would carry over from SK 1019. So asamāse "not in a compound" is said. Then api "also" is added, so that compounds like praṇāmaḥ still get ṇ.
In praṇinadati the ṇ of ni comes by SK 2285. The root's n stays n. It is separated from the r by the sounds of ni, which are not among those allowed to intervene (aṭ, a velar, a labial, āṅ, num; SK 197).
Roots listed after this rule. Root 55 ard "to go, to request". In liṭ, after doubling, SK 2248 lengthens the abhyāsa's a: ā ard a. The next rule works on this.
Rule. After that lengthened ā, the augment nuṭ is added to a root that has two consonants. nuṭ is marked with ṭ, so it attaches to the beginning of what follows. What is added is n. "That" refers to the a lengthened by SK 2248.
Example.
ard + ṇal → doubling and SK 2179: a ard a
→ SK 2248 lengthens the abhyāsa: ā ard a
→ ard has two consonants, r and d → nuṭ: ā n ard a
→ ānarda.
The aorist ārdīt has no abhyāsa, and so no nuṭ.
Roots listed after this rule.
pranardati has no ṇ. gard: gardati, jagarda.ānanta; andati, liṭ ānanda, with nuṭ by this rule.praṇindati (SK 2287).Watch out. nuṭ needs the long ā made by SK 2248. The aorist ārdīt has no abhyāsa and so no nuṭ. Where SK 2248 does not lengthen, nuṭ fails too: see āñcha under SK 2291.
Rule. The syllables ñi, ṭu and ḍu, when they stand at the beginning of a root as taught, are it-markers. "As taught" (upadeśe) and "it" are carried forward from SK 3; "root" from SK 18.
Example.
ṭunadi (root 67) → initial ṭu is an it-marker; the final i is also an it-marker
→ the root is nad, and it is idit → num (SK 2262) → nandati.
In the āśīrliṅ, yāsuṭ is kit, but an idit root keeps its n (SK 415): nandyāt.
Applies when / blocked by. One of Bhaṭṭoji's commentators gives the purpose of each marker. ñi allows the kta of ñītaḥ ktaḥ: suptaḥ. ṭu allows the affix athuc: nandathuḥ. ḍu allows the affix ktri: kṛtrimam.
Roots listed after this rule. The list is long; the useful forms are these.
śudhyāt.siseke. 87 aki: aṅkate, liṭ ānaṅke with nuṭ. 91 kuka: cukuke. 92 vṛka: varkate, liṭ vavṛke.ceke (e with loss of the abhyāsa).ṣvaṣkate, liṭ ṣaṣvaṣke.pranikakhati (kakh, where the n of ni stays); okhati, okhāṃcakāra.Two vārttikas come in this list.
vavṛke, not *vavarke.ṣvaṣkate keeps its ṣ. One of Bhaṭṭoji's commentators notes that the Mahābhāṣya rejects this vārttika; scholars discuss this.Watch out. Why is the vārttika needed for vavṛke? Guṇa is the later rule. Once guṇa gives var, the liṭ follows a conjunct and would no longer be kit. The vārttika settles it.
Rule. A final i-vowel or u-vowel of an abhyāsa is replaced by iyaṅ or uvaṅ when an unlike (asavarṇa) vowel follows. The ṅ is an it-marker: what comes is iy or uv. The words "i or u", "iyaṅ, uvaṅ" and "before a vowel" are carried forward from SK 271.
Example.
ukh + ṇal
→ doubling comes first (SK 2243), then SK 2179: u ukh a
→ guṇa of the root's light penultimate u: u okh a
→ the abhyāsa's u now stands before the unlike vowel o → uv: uv okh a
→ uvokha.
Applies when / blocked by. Why no ām? The guṇa vowel o gives an ic-initial root with a heavy vowel. By SK 2237 such a root takes ām in liṭ. But the guṇa was caused by ṇal. An operation caused by a combination cannot destroy that combination (the sannipāta-paribhāṣā). ām would cut the root off from ṇal. So no ām.
In the dual the ending atus is kit, so no guṇa:
u ukh atus → the abhyāsa's u stands before a like vowel, u → no uv
→ the two merge into ū → ūkhatuḥ, plural ūkhuḥ.
Should that ū be shortened, since it counts as part of the abhyāsa? No. The shortening of SK 2180 has already applied once. It applied first, like rain falling whether needed or not (parjanyavat). Bhaṭṭoji justifies the order by the maxim vārṇād āṅgaṃ balīyaḥ. It says an operation on the stem is stronger than one on sounds. He adds that the shortening is later. Bhaṭṭoji's commentators read "later" differently; scholars discuss this.
Roots listed after this rule.
uṅkhati. vakha: vavakhatuḥ (v-initial, so no e by SK 2263). makha: mekhatuḥ (e by SK 2260).tuṣṭuce (t kept by SK 2259).ānṛje.prārñjate (vṛddhi by SK 74); liṭ ṛñjāñcakre. 178 bhṛjī: bharjate, luṅ abharjiṣṭa.kucyāt. For 186 kruñca Bhaṭṭoji's text gives krucyāt. One of his commentators reads kruñcyāt instead, since its ñ is original; scholars discuss this.acyāt in the sense "to go", but añcyāt in the sense "to honour" (SK 424).Watch out. "Unlike" is the key word. Before a like vowel there is no uv, and the two vowels simply merge: ūkhatuḥ, not *uvūkhatuḥ.
aj → vivāya, vivyatuḥ; thal: vivayitha / vivetha / ājithaRoot 230 aj "to go" is replaced by vī before most ārdhadhātuka affixes. Its perfect raises a question: does thal take iṭ? Four rules and a summary verse settle when liṭ and thal take iṭ. The answer for vī is "optionally": vivayitha or vivetha. Where aj itself stays, it takes iṭ: ājitha.
Rule. vī replaces the root aj in the domain of an ārdhadhātuka affix. The affixes ghañ and ap are excepted. These are two affixes that make nouns from a root. "In the domain" means when such an affix is intended, before it is added.
Example.
aj + ṇal → vī replaces aj → vī + a
→ doubling: vī vī a → the abhyāsa is shortened: vi vī a
→ vṛddhi of the root's ī before ṇal: vi vai a → ai becomes āy: vivāya.
vī + atus → doubling, shortening → vi vī atus
→ atus is kit, so no guṇa → ī becomes y (SK 272) → vivyatuḥ; plural vivyuḥ.
Applies when / blocked by. A vārttika adds: before an ārdhadhātuka beginning with a val consonant, vī is optional. One of Bhaṭṭoji's commentators takes the rule's option as fixed by domain (a vyavasthitavibhāṣā). vī never comes with ghañ and ap, is optional before lyuṭ and val-initial affixes, and comes always elsewhere.
Why "in the domain"? So that vī can be made before yaṅ is added. Then the root begins with a consonant, and yaṅ can follow: vevīyate. Why not with ghañ and ap? samājaḥ, samajaḥ.
In vivyatuḥ the second v now stands before y, a consonant. Would SK 2265 lengthen the i before it? No. The y stands in place of the vowel ī, and by SK 50 it counts as that vowel. SK 51 does not stop this. A vārttika explains: in lengthening, only a deletion that replaces a vowel loses this status.
Watch out. What about thal? vī counts as anudātta, a root with a low-toned vowel. By SK 2246 such a root refuses iṭ. The next four rules take up the question.
Rule. After kṛ, sṛ, bhṛ, vṛ, stu, dru, sru and śru, the liṭ endings take no iṭ. "No" and "iṭ" are carried forward from SK 2981.
Example. Bhaṭṭoji gives no forms here. Bhaṭṭoji's commentators give:
tu stu + tha → no iṭ → tuṣṭotha.
tu stu + va → no iṭ → tuṣṭuva.
Applies when / blocked by. Bhaṭṭoji explains why the first four are named. It is a restriction (niyama). A bar on iṭ may rest on the root, as in SK 2246, or on the affix, as in SK 2381. In liṭ, any such bar holds only for kṛ, sṛ, bhṛ and vṛ. So every other root takes iṭ in liṭ. This is called the krādiniyama, "the restriction to kṛ and the rest". Bhaṭṭoji's commentators give examples: bibhidiva, babhūviva.
The other four, stu, dru, sru and śru, are named for two jobs. They block the iṭ that Bhāradvāja's view (SK 2296) would give in thal: tuṣṭotha. And they block the iṭ that the restriction itself would give before v and m: tuṣṭuva, tuṣṭuma.
Watch out. The restriction concerns only a bar on iṭ that comes from SK 2981 and its section. A root whose iṭ is merely optional, like ṣidhū, keeps its option. So siṣidhiva and siṣidhva both occur, Bhaṭṭoji's commentators say.
Rule. Take a root that ends in a vowel as taught, and that never takes iṭ before tās. After such a root, thal takes no iṭ either, just as tās does not. "As taught" is drawn back from SK 2295.
Example.
vī → luṭ vetā: vī never takes iṭ before tās
→ so thal too has no iṭ: vi vī + tha → guṇa → vivetha.
Bhaṭṭoji's commentators add cicetha and juhotha.
Applies when / blocked by. Bhaṭṭoji's commentators give a counter-example for each word:
bibheditha.jahartha from hṛ. After guṇa, har ends in a consonant; the rule still applies to hṛ as taught.sasvaritha from svṛ, which takes iṭ optionally (SK 2279).lulavitha from lū, which takes iṭ.papiva, papima.Rule. The same holds for a root that has short a in its taught form and never takes iṭ before tās. After it, thal takes no iṭ. Everything of SK 2294 except "vowel-final" is carried forward.
Example. Bhaṭṭoji's commentators give:
pac never takes iṭ before tās → thal without iṭ → papaktha.
Likewise śaśaktha from śak.
Applies when / blocked by. Counter-examples from Bhaṭṭoji's commentators. Why "as taught"? cakarṣitha. Why "short a"? bibheditha. Why short, not long? rarādhitha. Why "before tās"? jagrahitha. Why "never"? ānañjitha.
Rule. In the view of the grammarian Bhāradvāja, only a root ending in short ṛ refuses iṭ in thal. That is, among roots that never take iṭ before tās. Any other such root takes iṭ in thal. Set beside SK 2294–2295, Bhāradvāja's view gives an option.
Example.
vī + thal → with iṭ (Bhāradvāja): vi vay i tha → vivayitha
→ without iṭ (SK 2294): vivetha.
Where aj is not replaced (by the vārttika under SK 2292), aj takes iṭ: ājitha.
The rest of the paradigm of aj:
vivāya / vivaya (SK 2283); vivyiva, vivyima with iṭ by the krādiniyama.vetā / ajitā; lṛṭ veṣyati / ajiṣyati.vīyāt.Applies when / blocked by. Bhaṭṭoji sums up SK 2293–2296 in a verse:
Do stu, dru, sru and śru get the option too? No, says Bhaṭṭoji. Bhāradvāja's view cancels only the bar of the two rules just before it, SK 2294 and 2295. This follows the maxim anantarasya: a rule acts on what immediately precedes it. SK 2293 stands further back.
Bhaṭṭoji's commentators add examples of the option: papitha / papātha, pecitha / papaktha.
Watch out. A ṛ-final root like hṛ is aniṭ in thal by both views: jahartha. Bhāradvāja only adds iṭ to the other roots.
aj → avaiṣīt / ājīt; kṣi → kṣīyāt, akṣaiṣīt; kaṭe → akaṭīt; tip → tipsīṣṭa, atiptaThis group gathers the rules that shape the aorist and benedictive stems. SK 2291 adds more roots that may take aṅ; it stands before SK 2292 in the Kaumudī, but it belongs here in sense. SK 2297 gives vṛddhi to a final vowel before sic: avaiṣīt. SK 2298 lengthens a final vowel before y: kṣīyāt. SK 2299 lists roots that refuse sic-vṛddhi: akaṭīt. SK 2300 makes liṅ and sic kit in the ātmanepada, so there is no guṇa: tipsīṣṭa.
Rule. After jṝ, stambhu, mrucu, mlucu, grucu, glucu, gluñcu and śvi, aṅ optionally replaces cli. "cli" comes from SK 2222, "aṅ" from SK 2438, "optionally" from SK 2269.
Example.
a + mruc + cli + t
→ with aṅ: no guṇa → amrucat
→ without aṅ: sic with iṭ, guṇa, sic lost by SK 2266 → amrocīt.
Likewise amlucat / amlocīt, agrucat / agrocīt, aglucat / aglocīt.
gluñc: aglucat / agluñcīt.
Applies when / blocked by. In aglucat from gluñc the n is lost. One of Bhaṭṭoji's commentators reports that the author of the Vṛtti, a running commentary on the sūtras, keeps the n, since glucu and gluñcu are named separately. Scholars discuss this.
Roots listed after this rule.
sajjati; also ātmanepada sajjate.ānarca, with nuṭ.a āñch a. The t in SK 2248 (ata ādeḥ) takes in only a naturally short a. This a is not naturally short, so it is not lengthened, and no nuṭ comes. The vowels merge: āñcha. On the view that the t is only for ease of saying, nuṭ comes: ānāñcha.hūrchati, with ū by SK 2265.uñchāṃcakāra.Rule. An aṅga that ends in an ik vowel (i, u, ṛ) takes vṛddhi of that vowel before sic followed by a parasmaipada ending.
Example.
a + vī + s + īt (luṅ of aj with vī; no iṭ by SK 2246)
→ vṛddhi of ī: ai → a vai s īt → s becomes ṣ → avaiṣīt.
Without vī: āj + i + s + ī + t → ājīt.
lṛṅ: aveṣyat / ājiṣyat.
Applies when / blocked by. In ājīt the vṛddhi of SK 2267 is blocked by SK 2268, since the sic has iṭ. The long ā comes from āṭ merging with a. Bhaṭṭoji's commentators say the rule concerns only a final ik. So a consonant takes no vṛddhi, as in akoṣīt. One of them gives the counter-example for "parasmaipada": adhaviṣṭa.
Roots listed after this rule. 231 teja, 232 khaja, 233 khaji, 234 ejṛ (ejāṃcakāra), 235 ṭuosphūrjā (sphūrjati, pusphūrja). Then 236 kṣi "to waste away":
cikṣāya, cikṣiyatuḥ, cikṣiyuḥ;cikṣayitha / cikṣetha (the option of SK 2296);cikṣiyiva, cikṣiyima (iṭ by the krādiniyama);kṣetā.Rule. An aṅga ending in a vowel lengthens that vowel before an affix beginning with y. The affix must be neither a kṛt affix nor a sārvadhātuka affix. "y" is carried forward from SK 2649.
Example.
kṣi + yāt (āśīrliṅ; yāsuṭ is ārdhadhātuka) → i becomes ī → kṣīyāt.
luṅ by SK 2297: akṣaiṣīt.
Applies when / blocked by. Bhaṭṭoji's commentators give the counter-examples. A kṛt affix: prakṛtya. A sārvadhātuka affix: cinuyāt.
Roots listed after this rule.
avrājīt.ānaṭṭe (nuṭ). 256 ceṣṭa: aceṣṭiṣṭa. 261 aṭhi: aṇṭhate, ānaṇṭhe. 267 eṭha: eṭhāṃcakre.Rule. Before a sic that begins with iṭ, sic-vṛddhi does not come for these roots:
"vṛddhi" and "sic" come from SK 2297; "not" and "iṭ" from SK 2268.
Example.
a + kaṭ + i + s + ī + t (kaṭe is edit)
→ SK 2284 would allow vṛddhi → this rule blocks it
→ akaṭīt, never *akāṭīt.
Applies when / blocked by. One of Bhaṭṭoji's commentators gives an example for each item: agrahīt (h), avamīt (m), ahayīt (y), akṣaṇīt, aśvasīt, ajāgarīt, aśvayīt. Why "iṭ-initial sic"? adhākṣīt, from dah, has vṛddhi.
Roots listed after this rule.
peṭatuḥ (SK 2260).vavaṭatuḥ, vavaṭitha (v-initial: SK 2263 blocks e).iyāya, iyatuḥ, iyuḥ; thal iyayitha / iyetha. In iyāya the abhyāsa's i becomes iy before ā (SK 2290). Bhaṭṭoji's commentators derive iyatuḥ differently; scholars discuss this.asphuṭat / asphoṭīt.apaṭhīt / apāṭhīt (SK 2284; not edit).titipiṣe, with iṭ by the krādiniyama; luṭ teptā; lṛṭ tepsyate.Watch out. The rule blocks only vṛddhi before an iṭ-initial sic. kaṭī, with ī instead of e, is not edit. One of Bhaṭṭoji's commentators gives it both akaṭīt and akāṭīt.
Rule. Take a root ending in a consonant that stands next to an ik vowel. After it, a liṅ affix and a sic followed by an ātmanepada ending are kit, if they begin with a jhal. Being kit, they allow no guṇa. "ik" and "jhal" come from SK 2612, "consonant" from SK 2613, "kit" from SK 2242.
Example.
tip + sīṣṭa (āśīrliṅ)
→ the liṅ affix begins with s, a jhal, and follows p next to i → kit
→ no guṇa of i → tipsīṣṭa; tipsīyāstām, tipsīran.
luṅ: a + tip + s + ta → sic is kit, no guṇa
→ s lost between p and t (SK 2281) → atipta; atipsātām, atipsata.
Applies when / blocked by. "Followed by ātmanepada" qualifies only sic. The ātmanepada endings replace liṅ itself, so they cannot follow it. Bhaṭṭoji's commentators give the counter-examples. Why "ik"? yakṣīṣṭa, where kit would bring saṃprasāraṇa, the change of the ya of yaj to i. Why "ātmanepada"? adrākṣīt, where kit would block the augment am.
A sic with iṭ begins with i, not a jhal. So it is not kit by this rule.
Roots listed after this rule. 363 tepṛ: tepate, liṭ titepe. 364 ṣṭipṛ: liṭ tiṣṭipe (t kept by SK 2259). 365 ṣṭepṛ: tiṣṭepe. 366 glepṛ: glepate. 367 ṭuvepṛ: vepate. 374 trapūṣ "to be ashamed": trapate.
Watch out. Compare tepṛ, with e. Bhaṭṭoji's commentators note that titepe keeps its e. With i or u the kit liṭ would have blocked guṇa. That is why tepṛ is taught with e.
trap → trepe; jabh → jambhateTwo short rules extend earlier ones to named roots. SK 2301 gives the e-and-no-abhyāsa pattern of SK 2260 to four roots that would not qualify: trap gives trepe. SK 2302 gives num, the n of idit roots, to radh and jabh before a vowel: jambhate.
Rule. For tṝ, phal, bhaj and trap, the a becomes e and the abhyāsa is deleted. This happens before a kit liṭ ending and before thal with iṭ. The conditions are carried forward from SK 2260, 2471, 2363 and 2261.
Example.
trap + e (kit liṭ) → no reduplicated form survives
→ a becomes e → trepe; dual trepāte, plural trepire.
Applies when / blocked by. Bhaṭṭoji's commentators say why each root needs naming. The a of tṝ comes from guṇa, which SK 2263 would block. phal and bhaj have an abhyāsa with a changed sound. trap has a after a conjunct, not between single consonants. One of them gives the forms teratuḥ, phelatuḥ, bhejatuḥ.
trap is ūdit, so iṭ is optional (SK 2279): luṭ trapitā / traptā, āśīrliṅ trapiṣīṣṭa / trapsīṣṭa.
Roots listed after this rule.
ānambe (nuṭ by SK 2288).uttambhate, the s becoming like the d before it (SK 118). With vi: vistambhate, with no ṣ. SK 2272 takes only the sautra root, not ṣṭabhi. Mādhava finds the basis in the two spellings: SK 118 names the root with m, SK 2272 with n. Some say the ṭ of this root is original; for them, ṣṭambhate, liṭ ṭaṣṭambhe.Watch out. The rule reaches thal only with iṭ. It is the pair of SK 2260 and 2261 applied to four named roots.
Rule. The roots radh and jabh take the augment num when a vowel follows. "num" is carried forward from SK 2262. These roots are not idit, so SK 2262 alone would not give it.
Example.
jabh + a + te (the a of śap is a vowel) → num after the last vowel: ja-m-bh
→ jambhate. liṭ jajambhe; luṭ jambhitā (iṭ is a vowel); luṅ jambhiṣṭa.
Applies when / blocked by. Before a consonant there is no num. Bhaṭṭoji's commentators give raddhā and jabdham. jṛmbhate and jajṛmbhe get their n from jṛbhi's own i-marker, not from this rule.
Roots listed after this rule. 390 śalbha: śaśalbhe. 391 valbha: vavalbhe. 393 śrambhu: śrambhate, srambhate. 394 ṣṭubhu "to stop": stobhate; with vi viṣṭobhate (SK 2270); liṭ tuṣṭubhe; luṅ vyaṣṭobhiṣṭa (SK 2276). Then the parasmaipada roots begin with 395 gupū "to protect".
gup → gopāya → gopāyati, gopāyāṃcakāra / jugopa; kam → kāmi → kāmayate, kāmayāṃcakre / cakameSome roots take an affix in their own meaning before any verb ending. gup "to protect" takes āya and becomes gopāya. kam "to desire" takes ṇiṅ and becomes kāmi. These new forms are roots too (SK 2304). Before an ārdhadhātuka affix the extra affix is optional, so each root has two sets of forms. In liṭ the longer stem takes ām: gopāyāṃcakāra, kāmayāṃcakre. The bare root reduplicates: jugopa, cakame.
Rule. The affix āya comes after gupū, dhūpa, vicchi, paṇi and pani. It adds no meaning: it comes "in their own sense" (svārthe). "After a root" is carried forward from SK 2629.
Example.
gup + āya → guṇa of the light penultimate u (SK 2189) → gopāya.
Applies when / blocked by. āya comes after a root, so it is ārdhadhātuka. That is why guṇa applies, Bhaṭṭoji's commentators say.
Watch out. āya ends in a. That final a matters for the accent, one of Bhaṭṭoji's commentators notes, and it is deleted by SK 2308.
Rule. A form that ends in one of the affixes from san to ṇiṅ (SK 2310) is called a dhātu, a root.
Example.
gopāya ends in āya → it is a dhātu
→ it takes laṭ and the other lakāras → with śap: gopāya + a + ti → gopāyati.
Applies when / blocked by. Bhaṭṭoji's commentators list the twelve affixes: san, kyac, kāmyac, kyaṅ, kyaṣ, kvip, ṇic, yaṅ, yak, āya, īyaṅ and ṇiṅ. Of these, san is the desiderative affix. īyaṅ is the affix of the rule ṛter īyaṅ, read just before SK 2305. The rule says anta "ending in". In a naming rule, the mention of an affix does not by itself take in what ends in it. So the word is needed.
Watch out. gopāya is not in the dhātupāṭha. Without this rule it could not take verb endings.
Rule. The affixes āya, īyaṅ and ṇiṅ are optional when an ārdhadhātuka affix is intended.
Example.
gup, liṭ intended → with āya: gopāya + ām … → gopāyāṃcakāra
→ without āya: jugopa.
Applies when / blocked by. "When intended" reads the seventh case as one of domain (a viṣayasaptamī). Bhaṭṭoji's commentators say why. If it meant "when an ārdhadhātuka follows", āya would be added after the affix. Then in gopāyitā the base would not end in a when the ārdhadhātuka is taught, and SK 2308 could not delete the a.
Watch out. The rule covers only three affixes, āya, īyaṅ and ṇiṅ. They are the affixes of the three rules just before it in the Aṣṭādhyāyī, not all twelve of SK 2304.
Rule. In liṭ, the affix ām comes after the root kās and after a root that ends in an affix. This does not happen in a mantra.
Example.
gopāya + liṭ → gopāya ends in the affix āya → gopāya + ām.
Applies when / blocked by. A vārttika says: with kās, mention roots of more than one vowel (anekāc). Bhaṭṭoji explains: remove the word "affix" from the rule and put "of more than one vowel" in its place. So ām comes after kās and after any root of more than one vowel. Bhaṭṭoji's commentators give the reason: roots like cakās and jāgṛ also need ām. Some hold that ām comes even after a one-vowel root ending in an affix; scholars discuss this.
In a mantra there is no ām: one of Bhaṭṭoji's commentators gives kṛṣṇo nonāva.
Rule. This is a heading (adhikāra): "before an ārdhadhātuka affix". Its sense is read into the rules that follow, beginning with SK 2308. It is the second heading of this wording; the first is A 2.4.35 (SK 2432).
Example. A heading has no example of its own. In SK 2308, the a of gopāya is deleted before an ārdhadhātuka, ām.
Watch out. No commentator here states how far the heading runs. Treat it as governing the rules that need it, such as SK 2308 and SK 2313.
Rule. The a of a stem is deleted before an ārdhadhātuka affix. The stem must end in a at the moment the ārdhadhātuka affix is taught. "At the moment it is taught" (upadeśe) comes from SK 2428. The heading "ārdhadhātuka" is read twice: once with upadeśe, once as the following condition.
Example.
gopāya + ām → the a of gopāya is deleted → gopāy + ām
→ with the auxiliary: gopāyāṃcakāra, gopāyāṃbabhūva, gopāyām āsa.
Without āya (SK 2305): jugopa, jugupatuḥ.
Applies when / blocked by. gupū is ūdit, so iṭ is optional before thal (SK 2279): jugopitha / jugoptha. The other lakāras show up to three shapes: with āya, without āya but with iṭ, and without either:
gopāyitā / gopitā / goptā;gopāyyāt / gupyāt;agopāyīt / agopīt / agaupsīt.Why "when the ārdhadhātuka is taught"? Bhaṭṭoji's commentators give at and pat, from ay and pay with kvip. kvip is an affix that disappears entirely, leaving the root itself as the word. At that moment these stems end in y, so no a is deleted. Why "before an ārdhadhātuka"? So that in kathayati the deletion counts as still present (SK 50). The penultimate then takes no vṛddhi.
Roots listed after this rule.
tutumpatuḥ. A liṭ ending after a conjunct is not kit, so the n stays. The āśīrliṅ is kit (SK 2216), so the n goes: tupyāt.prastumpati gauḥ. It names the root with -ti, so the intensive with luk of yaṅ is excluded: pratotumpīti.sṛbhyāt.paṇāyati. In transacting: paṇate, liṭ peṇe. The ātmanepada marker has done its work on the bare root, so paṇāyati is parasmaipada. Likewise panāyati, panāyāṃcakāra, pene.cakṣamiṣe / cakṣaṃse, cakṣamidhve / cakṣandhve, cakṣamivahe.Watch out. The deleted a is the a of the stem gopāya, not the a of ām. Without the deletion, the vowels would merge anyway into gopāyām. One of Bhaṭṭoji's commentators says the deletion is still the proper route.
Rule. The m of a root ending in m becomes n when m or v follows. This is carried forward from SK 341, mo no dhātoḥ.
Example.
cakṣam + vahe (liṭ, uttama dual, without iṭ)
→ m before v becomes n → cakṣan vahe
→ n after ṣ becomes ṇ (SK 197) → cakṣaṇvahe.
Plural: cakṣamimahe with iṭ, cakṣaṇmahe without.
Applies when / blocked by. kṣamūṣ is ūdit, so each form comes with and without iṭ. Only the form without iṭ brings m next to m or v. Other forms: lṛṭ kṣamiṣyate / kṣaṃsyate; vidhiliṅ kṣameta; āśīrliṅ kṣamiṣīṣṭa / kṣaṃsīṣṭa; luṅ akṣamiṣṭa / akṣaṃsta.
Roots listed after this rule. 443 kamu, "in longing". Bhaṭṭoji glosses kānti here as desire.
Rule. After kam, the affix ṇiṅ comes in its own sense. ṇiṅ is heard as i. Its ṇ and ṅ are it-markers.
Example.
kam + ṇiṅ → kam + i
→ the penultimate a takes vṛddhi before a ṇit affix (SK 2282) → kāmi
→ kāmi is a root (SK 2304); ṇiṅ is marked with ṅ, so ātmanepada (taṅ) endings come
→ with śap, guṇa and ay: kāmayate.
Applies when / blocked by. One of Bhaṭṭoji's commentators explains why ṇiṅ carries the marker ṇ. It lets the ṇi-deletion of SK 2313 cover ṇiṅ as well as the causative ṇic. Without ṇiṅ, kam is still ātmanepada through its own marker: cakame.
Rule. ay replaces ṇi before ām, anta, ālu, āyya, itnu and iṣṇu. "ṇi" is carried forward from SK 2313. This is an exception to the deletion of ṇi taught in SK 2313.
Example.
kāmi + ām (liṭ; kāmi has more than one vowel, SK 2306)
→ ṇi would be deleted by SK 2313 → instead ay replaces it
→ kāmay + ām → with kṛ: kāmayāṃcakre.
Without ṇiṅ (SK 2305): cakame.
luṭ kāmayitā / kamitā; lṛṭ kāmayiṣyate / kamiṣyate.
Applies when / blocked by. One of Bhaṭṭoji's commentators gives one example for each affix: kārayām āsa (ām), gaṇḍayantaḥ (anta), spṛhayāluḥ (ālu), spṛhayāyyaḥ (āyya), stanayitnuḥ (itnu), pārayiṣṇavaḥ (iṣṇu).
Watch out. Would a bare ban on the deletion do, letting guṇa and ay follow? For kāmayāṃcakre, yes. One of Bhaṭṭoji's commentators says the substitute ay is needed where guṇa is not available, as in prastanayya.
kām i + luṅ → kām i a ta → kām a ta → kam a ta → ka kam a ta → ki kam a ta → kī kam a ta → acīkamataA root ending in ṇi, like kāmi, forms its aorist in a special way. The affix caṅ replaces cli. The ṇi is then deleted, the root vowel is shortened, and the root is reduplicated. The reduplicated syllable is treated as if before the desiderative affix san: it takes i and lengthens it. With the augment a, the result is acīkamata "he desired". This part gives the first rule; the chain continues with SK 2313.
Rule. After a root ending in ṇi, and after śri, dru and sru, the affix caṅ replaces cli. This happens when luṅ expresses the agent (kartṛ). "cli" comes from SK 2222 and "luṅ" from SK 2221. The c and ṅ of caṅ are it-markers; what comes is a.
Example.
kāmi + luṅ (ātmanepada, ta)
→ cli is added before the ending
→ caṅ replaces cli, instead of sic
→ kām i a ta.
Applies when / blocked by. This rule is an exception (apavāda) to SK 2222, which would give sic. By the convention for affixes, "ṇi" here means a root ending in ṇi, Bhaṭṭoji's commentators say. The rule covers kāmi, which ends in ṇiṅ, and the causatives in ṇic.
Watch out. kām i a ta is only a stage. The ṇi must still be deleted (SK 2313), and the root doubled, before you reach acīkamata.
kām i + luṅ → kām i a ta → kām a ta → kam a ta → ka kam a ta → ki kam a ta → kī kam a ta → acīkamata (continued)SK 2312 has put the affix caṅ in place of cli after a stem ending in ṇi. The rest of the group turns that string into a finished aorist. The running example is kāmi, the root kam "to desire" with the affix ṇiṅ, in luṅ, ātmanepada, third person singular. We start from kām i a ta, where a is what is left of caṅ. SK 2313 deletes the ṇi. SK 2314 shortens the ā. SK 2315 doubles kam. SK 2316 to SK 2318 then treat the first half, the abhyāsa, as if the desiderative affix san followed it. It gets i, then a long ī. The result is acīkamata "he desired".
Rule. The affix ṇi is deleted (lopa) when an ārdhadhātuka affix follows that does not begin with the augment iṭ. "ṇi" covers both ṇic and ṇiṅ, one of Bhaṭṭoji's commentators notes. The word lopa is carried forward from ato lopaḥ (SK 2308), and "ārdhadhātuka" comes from the heading.
Example.
kām i a ta → the a (from caṅ) is an ārdhadhātuka affix that does not begin with iṭ
→ yaṇ by SK 272 (i → y) also presents itself, and it is the later rule
→ a vārttika lets the deletion win instead
→ kām a ta.
Applies when / blocked by. Normally, when two rules conflict, the later one wins. Bhaṭṭoji therefore cites a vārttika: ṇyallopāv iyaṅyaṇguṇavṛddhidīrghebhyaḥ pūrvavipratiṣedhena. It means "deletion of ṇi and deletion of a win over iyaṅ, yaṇ, guṇa, vṛddhi and lengthening, by pūrvavipratiṣedha". Pūrvavipratiṣedha is the reverse of the usual order: in the conflict, the earlier rule wins.
Bhaṭṭoji's commentators both say how the vārttika is shared out. Deletion of ṇi wins over all five. Deletion of a wins only over vṛddhi and lengthening.
Why is the vārttika needed? Bhaṭṭoji answers: deletion of ṇi already has scope of its own. In pākti, from pācayati, none of the five competing operations arises. So, without the vārttika, the later rules would win wherever they did meet the deletion.
Bhaṭṭoji then gives his own conclusion. The rule says "aniṭi", "before an affix not beginning with iṭ". By the force of that wording, every such ārdhadhātuka affix is its domain. So the deletion is simply an exception (apavāda) to iyaṅ and the rest. An exception wins even over a later rule.
His five examples, one for each operation the deletion overrides:
iyaṅ: atatakṣat.
yaṇ: aṭiṭat.
guṇa: kāraṇā.
vṛddhi: kārakaḥ.
lengthening: kāryate.
One of Bhaṭṭoji's commentators derives them. atatakṣat is the causative of takṣ in luṅ with caṅ. kāryate is the passive of the causative of kṛ. The same commentator derives the yaṇ example as āṭiṭat, with the augment āṭ; scholars discuss this.
Watch out. The rule needs "not beginning with iṭ". Before an iṭ-initial affix, as in kārayitā, the ṇi stays and takes guṇa. Bhaṭṭoji's commentators add that the vārttika is still needed for deletion of a. They cite cikīrṣakaḥ and cikīrṣyāt.
Rule. Take an aṅga that stands before a ṇi which is itself followed by caṅ. Its penultimate vowel (upadhā) is replaced by a short vowel. "Of the aṅga" comes from the heading aṅgasya.
Example.
kām i a ta → SK 2313 deletes the ṇi → kām a ta
→ this rule shortens the penultimate ā of kām
→ kam a ta.
This is the order one of Bhaṭṭoji's commentators gives: first the deletion, then the shortening.
Applies when / blocked by. One of Bhaṭṭoji's commentators explains each word of the rule. "Penultimate": otherwise a form like acakāṅkṣat would be affected. "Before caṅ": without caṅ there is no shortening, as in kārayati. "Before ṇi": this keeps the rule off adīdapat, from dā with ṇi. There the ā is not the penultimate of the aṅga before ṇi, so the augment puk is not lost.
Watch out. The shortening works on the aṅga before ṇi, not on the root in general. In kām a ta the ṇi is already gone, yet the rule still gives kam a ta.
Rule. When caṅ follows, the first single-vowelled part (ekāc) of a root is doubled. That part must not already be an abhyāsa. For a root beginning with a vowel, the second single-vowelled part is doubled. The headings are SK 2175 and SK 2176. "Of a root that is not an abhyāsa" is carried from SK 2177, without its word liṭi.
Example.
kam a ta → caṅ follows, so kam is doubled
→ kam kam a ta
→ in the first half only the first consonant stays (halādiḥ śeṣaḥ, SK 2179)
→ ka kam a ta.
Applies when / blocked by. This is the same doubling you met in liṭ (SK 2177). Only the trigger differs: here it is caṅ. The first of the two halves is the abhyāsa (SK 2178).
Watch out. The rule does not double the ṇi-stem as a whole. It doubles the first single-vowelled part: here kam, not kāmi.
Rule. The abhyāsa is treated as if the affix san followed it (sanvat). Three things must hold:
Treatment "as if san followed" is called sanvadbhāva. Its effects are taught in SK 2317 and SK 2318.
Example.
ka kam a ta → the a of kam is a light vowel after the abhyāsa ka
→ after that a stood the ṇi, now deleted, which had caṅ after it
→ so ka is treated as if san followed.
One of Bhaṭṭoji's commentators notes that m stands between the light a and the ṇi. A single intervening sound does not count, by a maxim he cites.
Applies when / blocked by. Bhaṭṭoji says caṅpare is a bahuvrīhi: "that which has caṅ after it", namely the ṇi. Why must it be the ṇi? One of Bhaṭṭoji's commentators explains. The plain roots śri, dru and sru also take caṅ (SK 2312). If any sound before caṅ counted, aśiśriyat, adudruvat and asusruvat would be treated as before san. They would then wrongly get the lengthening of SK 2318.
Bhaṭṭoji gives two ways to construe the rule.
In ka kam a ta both construals give the same result. One of Bhaṭṭoji's commentators holds that only the first has the support of the Bhāṣya. He shows where they part: in the causative aorist of und, only the second allows the lengthening.
"Anaglope" is also needed. One of Bhaṭṭoji's commentators gives the counter-example acakathat, where an ak has been deleted.
Watch out. This rule does nothing by itself. It only says "treat the abhyāsa as before san". The next two rules say what that treatment does.
Rule. Before san, the a of the abhyāsa is replaced by i. "Of the abhyāsa" comes from SK 2620. "i" comes from SK 2496.
Example.
ka kam a ta → the abhyāsa ka is treated as before san (SK 2316)
→ its a becomes i
→ ki kam a ta.
Applies when / blocked by. The rule works in two ways. With a real san it gives desiderative forms. With sanvadbhāva, as here, it works in the caṅ-aorist. One of Bhaṭṭoji's commentators gives a desiderative example, pāpaciṣate. Whether the t of ataḥ is needed in the wording, scholars discuss this.
Watch out. Only an a of the abhyāsa changes. An abhyāsa with i or u is left to the next rule.
Rule. A light vowel of the abhyāsa is lengthened (dīrgha). This happens only where sanvadbhāva applies. The conditions of SK 2316 are carried forward, all except the word sanvat.
Example.
ki kam a ta → the light i of the abhyāsa is lengthened
→ kī kam a ta
→ the k of the abhyāsa becomes c, and the augment aṭ is added
→ acīkamata "he desired".
Applies when / blocked by. The affix ṇiṅ after kam is optional (SK 2305). Without it there is no ṇi, and so no caṅ by SK 2312. A vārttika fills the gap: kameś cleś caṅ vaktavyaḥ, "after kam, caṅ is to be stated in place of cli". Now there is still no ṇi. So neither sanvadbhāva nor the lengthening takes place. The form is acakamata, with a short a in the abhyāsa.
Five verses follow in Bhaṭṭoji's text. They say where the two operations reach in roots of more than one vowel.
acīcakāsat and acacakāsat.Watch out. The lengthening does not come from sanvadbhāva itself. One of Bhaṭṭoji's commentators is clear on this. It shares only the conditions of SK 2316. That is why Bhaṭṭoji names the two, "lengthening and sanvadbhāva", separately.
In the root list that follows (roots 444–464), vavaṇatuḥ keeps its abhyāsa and its a. The root vaṇ begins with v, which blocks the e-substitution (SK 2263). Compare senatuḥ, from ṣaṇ, where the e does come.
kram → krāmyati / krāmati, krāmyatu / krāmatu, akramītThis group runs through roots 465 to 480. First, jan, san and khan may take ā before a y-initial affix. Then three roots lengthen their vowel before a śit affix: an affix whose initial ś is an it-letter, such as śap. Eight roots, kram among them, may take the class-sign śyan instead of śap. Kram also lengthens before a śit affix in the parasmaipada. Last, a restriction settles when kram takes iṭ. The running example is kram "to step": krāmyati or krāmati "he steps", loṭ krāmyatu or krāmatu, and luṅ akramīt.
Rule. For jan, san and khan, the final n is optionally replaced by ā. This happens before a kit or ṅit affix that begins with y. Three words are carried forward: "of jan, san, khan" from SK 2504, "ā" from SK 2982, and "kit or ṅit" from SK 2428.
Example.
san (root ṣaṇa "to share"), āśīrliṅ: san + yāt
→ the n becomes ā: sa ā yāt
→ the two a-vowels join into a long ā
→ sāyāt
or, without the option, sanyāt.
Applies when / blocked by. One of Bhaṭṭoji's commentators introduces this rule as a special point for āśīrliṅ. The y must begin the affix; the word ye is read that way. Later Bhaṭṭoji uses the rule for khan: khāyāt or khanyāt (SK 2363).
In the root list that follows, dram "to go" gives luṅ adramīt with no vṛddhi, by SK 2299. Am gives amati and liṭ āma. Mīmṛ gives liṭ mimīma.
Rule. The vowel of ṣṭhiv, klam and cam is lengthened before a śit affix. "Lengthened" is carried forward from SK 2519.
Example.
ā + cam + śap + ti → before the śit affix śap, the a of cam is lengthened
→ ācāmati "he sips".
One of Bhaṭṭoji's commentators adds ṣṭhīvati "he spits" and klāmyati "he is weary".
Applies when / blocked by. A vārttika limits cam: āṅi cama iti vaktavyam, "for cam, only after āṅ". Without āṅ there is no lengthening: camati, vicamati. Bhaṭṭoji also lists luṅ acamīt. It has no vṛddhi, by SK 2299, one of his commentators notes. One of Bhaṭṭoji's commentators reads the rule itself as naming ācam; scholars discuss this.
Some also read a root jimi here, giving jemati, Bhaṭṭoji notes.
Watch out. The rule needs a śit affix. Later rules take over the word śiti from here (SK 2322, SK 2346, SK 2360).
Rule. After eight roots, śyan optionally comes before a sārvadhātuka affix in the sense of the agent. The roots are bhrāś, bhlāś, bhram, kram, klam, tras, truṭ and laṣ. Śyan is the class-sign of the fourth class, the divādi roots (SK 2505). "In the sense of the agent" comes from SK 2167, and "sārvadhātuka" from SK 2756.
Example.
kram + ti → with śyan: kram + ya + ti
→ without it, the usual śap: kram + a + ti.
With SK 2322 these give krāmyati and krāmati.
Applies when / blocked by. One of Bhaṭṭoji's commentators calls this an option both ways. Klam and tras are divādi roots, so for them śyan would otherwise be obligatory. For the others śap would otherwise be obligatory. Bhaṭṭoji uses the rule again for laṣ (laṣyati, laṣati, SK 2363), for bhram (bhramyati, bhramati, SK 2355) and for bhrāś (bhrāśyate, bhrāśate, SK 2354).
Rule. The vowel of kram is lengthened before a śit affix that has a parasmaipada ending after it. "Lengthened" comes from SK 2519, and "before a śit affix" from SK 2320.
Example.
kram + ya + ti → śyan is śit, and ti is parasmaipada → krāmyati.
kram + a + ti → śap is śit → krāmati.
loṭ: krāmyatu, krāmatu.
Applies when / blocked by. Why "in the parasmaipada"? One of Bhaṭṭoji's commentators answers with upakramate and parākramate. There the ending is ātmanepada, and the vowel stays short.
Watch out. Bhaṭṭoji also lists liṭ cakrāma. Its long ā is not from this rule, since liṭ has no śit affix. It is the vṛddhi of the penultimate before ṇal.
Another trap: loṭ second person singular, where the ending hi is deleted by luk. The form is krāma, still lengthened. One of Bhaṭṭoji's commentators explains. The lengthening is caused by the śit affix śap, not by the deleted ending.
Rule. After snu and kram, a val-initial ārdhadhātuka affix takes iṭ only when there is no cause of ātmanepada. The iṭ is the one of SK 2184.
Example.
kram, luṅ, parasmaipada: nothing calls for ātmanepada → iṭ comes → akramīt "he stepped".
One of Bhaṭṭoji's commentators notes: no vṛddhi here, by SK 2299.
Applies when / blocked by. Snu and kram would take iṭ anyway. So this rule is a restriction (niyama): "only here". Bhaṭṭoji says just that: "only here is there iṭ". A niyama does not add an operation; it narrows one already available.
Where a cause of ātmanepada is present, there is no iṭ. Bhaṭṭoji's commentators give upakraṃsyate and upasnoṣyate jalena. Why "cause" (nimitta), and not simply "before an ātmanepada ending"? One of Bhaṭṭoji's commentators answers. Forms like cikramiṣiṣyati and pracikraṃsiṣyate would otherwise come out wrong.
The next roots, ay and six others (roots 474–480), are anudātteta: they take ātmanepada. ayate "he goes".
ay → ayāṃcakre, āyiḍhvam / āyidhvam, palāyate; pyāy → pipye, apyāyiThis group follows the ātmanepada roots ay "to go" to tāy (roots 474–489). Ay forms its liṭ with ām, like day and ās. After iṭ, a dh in some endings may turn retroflex. Before ay, the r of a prefix becomes l: palāyate "he runs away". The root pyāy "to grow" becomes pī in liṭ: pipye. And after some roots the aorist may take ciṇ, which removes the ending ta: apyāyi.
Rule. After day, ay and ās, the affix ām comes in liṭ. "Ām in liṭ" is carried forward from SK 2306, which gave ām after kās and after roots ending in an affix.
Example.
ay + liṭ → ām is added: ay + ām
→ liṭ of kṛ follows as the auxiliary, as in edhāṃcakre earlier in this chapter
→ ayāṃcakre "he went".
Bhaṭṭoji adds two more forms of ay: vidhiliṅ ayeta, and āśīrliṅ ayiṣīṣṭa.
Applies when / blocked by. The same rule gives dayāṃcakre from day "to give" (root 481), in the list after SK 2326.
Rule. Take the dh of ṣīdhvam (āśīrliṅ), and of the dh-initial endings of luṅ and liṭ. It is optionally replaced by a retroflex, ḍh. This happens when these endings come after an iṭ that follows a sound of the group iṇ. The words are carried forward from SK 2247, where the retroflex is obligatory. "A retroflex" comes from SK 210.
Example.
āśīrliṅ, second person plural: ay + iṭ + ṣīdhvam → ayiṣīdhvam
→ the dh of ṣīdhvam may become ḍh
→ ayiṣīḍhvam or ayiṣīdhvam.
luṅ: third person singular āyiṣṭa; second person plural āyiḍhvam or āyidhvam.
Applies when / blocked by. One of Bhaṭṭoji's commentators derives the luṅ forms. Sic replaces cli, the augment āṭ comes with vṛddhi, and the s of sic is lost before dhvam by SK 2249. The other commentator gives a liṭ pair: lilihiḍhve and lilihidhve.
Bhaṭṭoji uses the rule again in the root list after SK 2334. There gāh gives liṭ jagāhiḍhve or jagāhidhve.
Rule. When the root ay follows, the r of an upasarga (a verbal prefix) is replaced by l. The words "r … l" are carried forward from kṛpo ro laḥ, SK 2350. That rule is taught later in the Kaumudī, but it stands earlier in the Aṣṭādhyāyī.
Example.
pra + ayate → the r of pra becomes l
→ plāyate "he goes forth".
parā + ayate → palāyate "he runs away".
Applies when / blocked by. What of nis and dus? Their final s becomes ru before ay. But that ru counts as not yet made (asiddha) for this rule. So there is no l: nirayate, durayate. The prefixes nir and dur are different: they end in a real r. They give nilayate, dulayate.
The word pratyaya does not get l. Bhaṭṭoji explains: it comes from prati with the root i (iṇ), not from ay.
The poet Māgha writes udayati, a parasmaipada participle of ud + ay. But ay is an ātmanepada root, so how? Bhaṭṭoji gives two answers.
In the root list that follows, vay gives liṭ vavaye. It begins with v, so there is no e and no loss of the abhyāsa. The other roots give peye, meye, ceye, teye. Ṇay gives praṇayate. Ūyī gives ūyāṃcakre.
Rule. In liṭ and before the intensive affix yaṅ, the root pyāy is replaced by pī. "pyāy → pī" is carried forward from SK 3072.
Example.
opyāyī "to grow", liṭ ātmanepada: pyāy + e
→ pyāy is replaced by pī: pī + e
→ doubling applies again: pī pī e
→ the vowel of the abhyāsa is short (SK 2180): pi pī e
→ yaṇ by SK 272 (ī → y)
→ pipye "it grew".
Dual pipyāte, plural pipyire.
For yaṅ, one of Bhaṭṭoji's commentators gives pepīyate.
Applies when / blocked by. The substitution pī is the later rule, so it goes first and blocks the doubling. Can doubling come back? Normally what is blocked in a conflict stays blocked. Bhaṭṭoji uses a principle, punaḥprasaṅgavijñāna: a rule blocked in a conflict may apply again once the other rule has applied. So the doubling returns after pī. Some explain the doubling differently; scholars discuss this.
Watch out. Bhaṭṭoji names only three steps: pī, the renewed doubling, and yaṇ. The short vowel of the abhyāsa is the general rule for every abhyāsa.
Rule. After dīp, jan, budh, pūr, tāy and pyāy, the affix ciṇ optionally replaces cli. This happens when the ending ta follows. "Cli" comes from SK 2222, and "ciṇ when ta follows" from SK 2513. Ciṇ is heard as i.
Example.
a + pyāy + cli + ta → ciṇ replaces cli: a + pyāy + i + ta
→ SK 2329 removes the ta
→ apyāyi "it grew".
Without ciṇ, sic comes: apyāyiṣṭa.
Applies when / blocked by. Bhaṭṭoji says "the singular ta". One of Bhaṭṭoji's commentators explains why: these roots take ātmanepada, so a ta after them is always the third person singular ending.
One of Bhaṭṭoji's commentators identifies dīp, jan, budh and pūr as divādi roots. Bhaṭṭoji makes the same point after SK 2362. The bhvādi root budhir does not take this ciṇ. The roots before and after it in the rule are divādi, so only the divādi budh is meant.
Rule. What comes after ciṇ is deleted by luk. One of Bhaṭṭoji's commentators says what that is: only the ending ta.
Example.
a + pyāy + i + ta → the ta after ciṇ is deleted
→ apyāyi.
tāy "to extend, to protect": atāyi with ciṇ, atāyiṣṭa without.
Applies when / blocked by. The other commentator adds that naming ta in the gloss only states what is implied. The deletion falls on the whole affix.
In the root list that follows (roots 489–515), hay "to go" gives luṅ ahayīt. There is no vṛddhi, because the root ends in y. Ṣev "to serve" gives pariṣevate, with ṣ by SK 2275, and liṭ siṣeve. Tev and dev "to sport" give titeve and dideve.
dhinv → dhin + a + u → dhinoti; dhinvaḥ / dhinuvaḥ; dhinuThis group covers roots 515 to 649. First, an a before a final l or r takes vṛddhi in the parasmaipada sic-aorist: ālīt. Then the j of ji "to conquer" becomes g after its abhyāsa: jigāya. Last, two roots, dhinv "to please" and kṛṇv "to injure", take the affix u instead of plain śap. The running example is dhinv. Its v becomes a, u is added, and the a is lost: dhinoti "he pleases". Before m and v the u may be lost: dhinvaḥ or dhinuvaḥ. In the imperative the ending hi is dropped: dhinu.
Rule. Take an aṅga that ends in l or r, where that l or r stands next to an a. That a takes vṛddhi before sic followed by a parasmaipada ending. "Vṛddhi before sic in the parasmaipada" is carried forward from SK 2297.
Example.
al "to adorn", luṅ parasmaipada with sic, after mā
→ the a next to the final l takes vṛddhi
→ mā bhavān ālīt "may you not adorn".
Others in the root list: aphālīt (phal), askhālīt (skhal), agālīt (gal), atsārīt (tsar), aśvālīt (śval).
Applies when / blocked by. This rule is an exception to two others, Bhaṭṭoji says. SK 2268 (neṭi) forbids the vṛddhi of a consonant-final root before a sic with iṭ. SK 2284 (ato halādeḥ…) makes it optional. Here the vṛddhi is obligatory.
Counter-example: abhr "to go" gives mā bhavān abhrīt, with no vṛddhi. The final r does not stand next to an a. One of Bhaṭṭoji's commentators adds akhorīt and amīlīt, which have no a at all.
Why "mā bhavān"? One of Bhaṭṭoji's commentators explains, in a later passage. With mā the augment is not added. With the augment āṭ, the form would look the same with or without vṛddhi.
Bhaṭṭoji explains the wording. Lra is "l or r" with its sixth-case ending elided. Anta here means "near", not "end". And ataḥ is taken twice: once with "near", once as the vowel that takes vṛddhi. Bhaṭṭoji's commentators both give this reading.
In the root list, phal gives liṭ phelatuḥ, pheluḥ by SK 2301. Ṣṭhiv gives ṣṭhīvati by SK 2320.
Watch out. The vṛddhi falls on the a next to the final consonant, not on the final consonant. Without "ataḥ" taken twice, the rule would give vṛddhi to the last sound, one of Bhaṭṭoji's commentators warns.
Rule. An abhyāsa may be caused by san or by liṭ. After such an abhyāsa, the j of ji "to conquer" is replaced by a velar. "After the abhyāsa" comes from SK 2430, and "a velar" from SK 2863.
Example.
ji + ṇal → doubling: ji ji a
→ vṛddhi of the second i before ṇal (ai, which becomes āy before the a): ji jāy a
→ the j after the abhyāsa becomes the velar g
→ jigāya "he conquered".
Dual jigyatuḥ, plural jigyuḥ. Second person singular jigayitha or jigetha. First person singular jigāya or jigaya. First person dual and plural jigyiva, jigyima.
Applies when / blocked by. Where there is no abhyāsa, the j stays. luṭ jetā, āśīrliṅ jīyāt, luṅ ajaiṣīt.
"Caused by san or liṭ" matters. Bhaṭṭoji's commentators, following Mādhava, give jejayiṣati. That is san added to an intensive stem. Its abhyāsa was not caused by the san, so the j stays.
In the root list (roots 562–594), urvī gives ūrvāṃcakāra. Turvī gives tutūrva, with the long vowel by SK 2265. Ivi gives invati and invāṃcakāra.
Rule. Where śap would come, two things happen to dhinv and kṛṇv. Their final v is replaced by a. And the affix u is added. The u is drawn in by "ca" from SK 2466. The conditions "for the agent" and "before a sārvadhātuka" come from SK 2167 and SK 2756.
Example.
dhinv + ti → v becomes a, and u is added: dhin a u ti
→ the a is lost by ato lopaḥ (SK 2308): dhin u ti
→ the u takes guṇa before the p-marked ending tip
→ dhinoti "he pleases".
Dual dhinutaḥ, plural dhinvanti.
Applies when / blocked by. Why does the i of dhin not take guṇa before the u? The lost a still counts as present (sthānivat). So the i is not treated as the penultimate. Bhaṭṭoji says so.
One of Bhaṭṭoji's commentators asks: why make v into a and then delete it? Why not delete v at once? A deleted v is not the substitute of a vowel. So it could not count as present, and the i would take guṇa.
Dhinv is dhivi with the nasal num, and kṛṇv is kṛvi with num. Bhaṭṭoji says kṛṇoti goes like dhinoti.
Watch out. The a and the u are taught together. Still the a comes first. So when the u arrives, the stem ends in a, and ato lopaḥ can apply. Bhaṭṭoji's commentators both explain this order.
Rule. Take an aṅga ending in a u that belongs to an affix and is not preceded by a conjunct. That u is optionally deleted before m or v. The words "such a u" refer to the next rule, SK 2334, which stands earlier in the Aṣṭādhyāyī.
Example.
dhinu + vas → the u may be lost before v
→ dhinvaḥ or dhinuvaḥ "we two please".
dhinu + mas → dhinmaḥ or dhinumaḥ "we please".
Applies when / blocked by. Before mip, the first person singular ending, guṇa wins because it is the later rule: dhinomi.
One of Bhaṭṭoji's commentators gives counter-examples. In yuvaḥ, yumaḥ the u is not part of an affix. In śaknuvaḥ, śaknumaḥ a conjunct stands before the u. He also notes that "a u belonging to an affix" covers śnu, as in sunumaḥ.
Rule. Take an aṅga ending in a u that belongs to an affix and has no conjunct before it. After it, the imperative ending hi is deleted by luk. "Luk" comes from SK 2329, and "of hi" from SK 2202.
Example.
dhinu + hi → the hi is deleted
→ dhinu "please!".
Applies when / blocked by. In the first person of loṭ, the augment āṭ comes (SK 2204). It is nitya: it applies whether or not the u is lost. So it comes first, before the u can be deleted by SK 2333. Then the u takes guṇa and becomes av: dhinavāva, dhinavāma "let us two please", "let us please".
A conjunct before the u blocks the rule. So akṣ with śnu keeps its ending: akṣṇuhi (SK 2338).
In the root list (roots 595–649), kṛvi gives kṛṇoti, like dhinoti. Av gives mā bhavān avīt, where neṭi (SK 2268) blocks vṛddhi. Dhāvu has a svarita it-vowel, so it takes both padas: dhāvati, dhāvate. Then come the ātmanepada roots ending in a sibilant or h, the ūṣman sounds. Īkṣ gives īkṣāṃcakre, and gāh gives liṭ jagāhe, jaghākṣe, jaghāḍhve.
gṛh → aghṛkṣata, aghṛkṣātām; akṣ → akṣṇoti / akṣati; ghas → ghatsyati, aghasatThe roots of this group end in an ūṣman sound: a sibilant (ś, ṣ, s) or h. Their endings meet many consonant rules. One ḍh is lost before another. The aorist of some of them takes the affix ksa in place of cli. Akṣ and takṣ may take the class-sign śnu. A few roots take iṭ optionally before t, and uṣ takes ām optionally in liṭ. The s of ghas becomes t before s. Last, three sets of roots take the aorist affix aṅ. The running examples: gṛh "to censure" gives aghṛkṣata, aghṛkṣātām. Akṣ "to pervade" gives akṣṇoti or akṣati. Ghas "to eat" gives ghatsyati and aghasat.
Rule. A ḍh is deleted when a ḍh follows. The first of the two goes.
Example.
gāh "to stir", luṭ without iṭ: gāh + tā
→ h becomes ḍh (SK 324): gāḍh + tā
→ t becomes dh after it (SK 2280): gāḍh + dhā
→ dh becomes ḍh by ṣṭutva, the retroflex assimilation (SK 113): gāḍh + ḍhā
→ this rule deletes the first ḍh
→ gāḍhā.
This chain is the one Bhaṭṭoji's commentators give. Bhaṭṭoji lists the forms of gāh: gāḍhā; luṅ agāḍha, agāḍhāḥ, aghāḍhvam. With iṭ: agāhiṣṭa.
Applies when / blocked by. One of Bhaṭṭoji's commentators notes that ṣṭutva is not treated as unmade (asiddha) for this rule. Otherwise the rule could never apply.
Next comes the root gṛhū "to censure" (root 650). Its liṭ is jagṛhe, with no guṇa. A vārttika decides it: ṛdupadhebhyo liṭaḥ kittvaṃ guṇāt pūrvavipratiṣedhena. "For roots with ṛ as penultimate, the kit-status of liṭ wins over guṇa, by pūrvavipratiṣedha." One of Bhaṭṭoji's commentators names the rival: guṇa by SK 2189, the later rule. More forms: jagṛhiṣe or jaghṛkṣe; jaghṛḍhve; luṭ garhitā or garḍhā; lṛṭ garhiṣyate or gharkṣyate.
Watch out. The rule needs a following ḍh. The ḍh of gāḍh is lost only after the t has become ḍh too.
Rule. The affix ksa replaces cli after a root with three features. It ends in a sound of the group śal, the ūṣman sounds. Its penultimate is a vowel of the group ik (i, u, ṛ, ḷ). And it takes no iṭ (aniṭ). The k of ksa is an it-letter, so the affix is kit.
Example.
gṛh, luṅ ātmanepada, without iṭ: a + gṛh + ksa + ta
→ ksa is kit, so the ṛ takes no guṇa
→ h becomes ḍh, the initial g becomes gh, ḍh becomes k, and the s becomes ṣ
→ aghṛkṣata "he censured".
The steps are those given by one of Bhaṭṭoji's commentators.
Applies when / blocked by. This rule is an exception to sic. One of Bhaṭṭoji's commentators tests each condition. Not ending in śal: atipta. No ik penultimate: agāḍha. Taking iṭ: auhiṣṭa. None of these takes ksa.
Bhaṭṭoji uses the rule often in this part. Kruś gives akrukṣat, ruh gives arukṣat (SK 2362). Śiṣ gives aśikṣat (SK 2339). Viṣ gives avikṣat (SK 2341). Guh gives aghukṣat (SK 2364).
Watch out. The root must be aniṭ in the form at hand. Gṛh and guh take iṭ optionally. With iṭ they take sic; only without it do they take ksa.
Rule. Before a taṅ ending beginning with a vowel, ksa is deleted. By SK 42 only its final sound, the a, goes. "Deletion" is carried forward from SK 3584.
Example.
aghṛkṣa + ātām → the a of ksa is deleted before the vowel-initial ātām
→ aghṛkṣātām "they two censured".
Plural: aghṛkṣanta.
Applies when / blocked by. Why must the a go? One of Bhaṭṭoji's commentators explains. If it stayed, SK 2235 (āto ṅitaḥ) would replace the ā of ātām with iy. With the a gone, the ā no longer follows an a.
The plural ending jha does not begin with a vowel, so the rule does not touch it at first. It becomes anta, giving aghṛkṣanta.
Bhaṭṭoji glosses the rule with "before a taṅ ending". Some call the word taṅ a later addition; others draw "ātmanepada" back from the following rule, SK 2365. Scholars discuss this.
In the root list, ghuṣir "to proclaim" has ir as it-marker. So it takes aṅ optionally (SK 2269): aghuṣat or aghoṣīt.
Rule. After akṣ "to pervade", the affix śnu optionally comes before a sārvadhātuka affix in the sense of the agent. Otherwise śap comes. Śnu is the class-sign of the fifth class (SK 2523).
Example.
akṣ + ti → with śnu: akṣ + nu + ti
→ the u takes guṇa before tip, and n becomes ṇ
→ akṣṇoti "he pervades".
Dual akṣṇutaḥ, plural akṣṇuvanti. With śap: akṣati, akṣataḥ, akṣanti.
Applies when / blocked by. Bhaṭṭoji gives the other lakāras too.
ānakṣa. Second person singular ānakṣitha or ānaṣṭha: akṣ has ū as it-letter, so iṭ is optional.akṣitā or aṣṭā.akṣiṣyati, or without iṭ akṣyati. There the k is lost by SK 380, and ṣ becomes k before s by SK 295.akṣṇotu, akṣṇuhi, akṣṇavāni.ākṣṇot, first person singular ākṣṇavam.akṣṇuyāt; āśīrliṅ akṣyāt.In akṣṇuhi the hi stays. A conjunct stands before the u, so SK 2334 does not apply.
luṅ has two sets. With iṭ, SK 2268 blocks the vṛddhi: mā bhavān akṣīt, akṣiṣṭām, akṣiṣuḥ. Without iṭ, vṛddhi comes: mā bhavān ākṣīt, āṣṭām, ākṣuḥ.
Watch out. Śnu and śap belong only to the sārvadhātuka lakāras. In āśīrliṅ, which is ārdhadhātuka, there is neither: akṣyāt.
Rule. After takṣ in the sense of making thin (tanūkaraṇa), śnu optionally comes where śap would come.
Example.
takṣ + ti → with śnu: takṣṇoti kāṣṭham "he planes the wood"
→ with śap: takṣati kāṣṭham.
liṭ tatakṣitha or tataṣṭha. luṅ with iṭ atakṣīt, atakṣiṣṭām; without iṭ atākṣīt, atāṣṭām.
Applies when / blocked by. Why "in the sense of making thin"? Bhaṭṭoji answers: vāgbhiḥ saṃtakṣati "he pares with words", that is, "he reviles". There only śap comes.
One of Bhaṭṭoji's commentators draws a lesson. Takṣū (root 655) is listed only with the sense "making thin", yet the rule has to name that sense. So the meanings in the dhātupāṭha are only indicative (upalakṣaṇa). A root may be used in other senses too.
In the root list (roots 657–694), ukṣ gives ukṣāṃcakāra. Śiṣ gives luṭ śeṣṭā and luṅ aśikṣat, with ksa (SK 2336). Vaṣ gives vavaṣatuḥ, keeping its abhyāsa by SK 2263.
Rule. After iṣ "to wish", sah, lubh, ruṣ and riṣ, an ārdhadhātuka affix beginning with t optionally takes iṭ.
Example.
ruṣ "to injure", luṭ: ruṣ + tā
→ with iṭ: roṣitā
→ without iṭ, t becomes ṭ after ṣ: roṣṭā.
riṣ: reṣitā or reṣṭā.
Applies when / blocked by. The option is only before t. In lṛṭ the affix sya begins with s, so iṭ is not optional: roṣiṣyati, reṣiṣyati.
One of Bhaṭṭoji's commentators gives the other roots: eṣitā or eṣṭā, sahitā or soḍhā, lobhitā or lobdhā. The other cites a vārttika: iṣ of the fourth class, with śyan, is not included.
Watch out. "Iṣ" in the rule is icchati, the root "to wish". Bhaṭṭoji glosses it that way.
Rule. After uṣ "to burn", vid and jāgṛ, ām optionally comes in liṭ. "Ām in liṭ" comes from SK 2306.
Example.
uṣ, liṭ: with ām: oṣāṃcakāra
→ without ām: uvoṣa, dual ūṣatuḥ, second person singular uvoṣitha.
In uvoṣa the abhyāsa u becomes uv by SK 2290, Bhaṭṭoji's commentators note.
Applies when / blocked by. The root list after this rule is rich in rules already met.
viveṣitha, viviṣiva. Here iṭ comes by the krādi restriction (SK 2293), which gives iṭ in liṭ to all roots except kṛ and a few others. luṭ veṣṭā, lṛṭ vekṣyati, luṅ avikṣat.poṣitā, poṣiṣyati, apoṣīt. The list of aniṭ roots names puṣya, with śyan, so this puṣ takes iṭ. Hence no ksa. And the aṅ of SK 2343 is for the fourth-class puṣ, so no aṅ either.śleṣitā. Kaiyaṭa and others hold that only the fourth-class śliṣ is aniṭ.Ghas "to eat" is not used in every form, Bhaṭṭoji says. SK 2424 teaches ghas as an optional substitute for ad in liṭ, and that would be pointless otherwise. So ghas is used only where a sign (liṅga) or a direct statement allows it.
ghasati.ghastā.Rule. An s is replaced by t before an ārdhadhātuka affix beginning with s. "t" is carried forward from SK 3078.
Example.
ghas, lṛṭ: ghas + sya + ti
→ the s of ghas becomes t before sya
→ ghatsyati "he will eat".
Bhaṭṭoji adds loṭ ghasatu, laṅ aghasat, vidhiliṅ ghaset.
Applies when / blocked by. One of Bhaṭṭoji's commentators gives counter-examples. ghasmaraḥ: the affix does not begin with s. vaste: the affix is not ārdhadhātuka.
Ghas has no āśīrliṅ, Bhaṭṭoji says, since nothing indicates that use (see SK 2341).
Rule. Before parasmaipada endings, the affix aṅ replaces cli after three sets of roots:
"Cli" comes from SK 2222, and "aṅ" from SK 2438.
Example.
ghasḷ, luṅ: a + ghas + cli + t
→ the root has ḷ as it-letter, so aṅ replaces cli: a + ghas + a + t
→ aghasat "he ate".
Applies when / blocked by. Why the fourth-class puṣ? A root puṣ is found in four classes. Bhaṭṭoji's commentators reach the fourth by elimination. The first-class group would make "dyut and the rest" redundant, since dyut follows it directly. The ninth-class group has only four roots, which could simply have been marked with ḷ. In the tenth class, ṇic stands before cli.
In the root list (roots 716–741):
ahasīt.śeśatuḥ, śeśuḥ, śeśitha, with e and loss of the abhyāsa.śaśasatuḥ, śaśasuḥ.śasyāt.atuhat or atohīt.mā bhavān uhat or auhīt.ānarha.Then come the anudātteta roots up to kṛp, beginning with dyut: dyotate.
Watch out. Aṅ comes only with parasmaipada endings. With ātmanepada endings these roots take sic, as the next group shows.
dyut → didyute, adyutat / adyotiṣṭa; vṛt → vartsyati / vartiṣyate, avṛtat / avartiṣṭa; kṛp → kalpate, kalptāsiThe dyutādi are ātmanepada roots, from dyut "to shine" to kṛp (roots 741–762). Some of them may take parasmaipada in certain forms. Dyut keeps a vowel in its abhyāsa: didyute. In luṅ all of them may take parasmaipada, and then aṅ: adyutat or adyotiṣṭa. Five of them, the vṛtādi, may take parasmaipada before sya and san. Four of these then refuse iṭ: vartsyati beside vartiṣyate. Syand may take ṣ after some prefixes. Kṛp has l for its r: kalpate. It may also take parasmaipada in luṭ, and then no iṭ: kalptāsi.
Rule. The abhyāsa of dyut and of svāpi takes saṃprasāraṇa: its semivowel is replaced by the matching vowel. Svāpi is svap with the causative ṇic, one of Bhaṭṭoji's commentators explains. "Of the abhyāsa" comes from SK 2620.
Example.
dyut, liṭ ātmanepada: dyut dyut e
→ the abhyāsa loses its final t, but this rule saves its y: dyu dyut e
→ the y becomes i (saṃprasāraṇa): diu dyut e
→ i + u gives i by pūrvarūpa (SK 330): di dyut e
→ didyute "it shone".
Dual didyutāte. luṭ dyotitā.
Applies when / blocked by. Without this rule, halādiḥ śeṣaḥ (SK 2179) would drop the y of the abhyāsa. So this rule is an exception to it, one of Bhaṭṭoji's commentators says.
For svāpi the other commentator gives suṣvāpayiṣati. He adds a condition: the ṇic-stem must stand right before the affix that causes the abhyāsa.
Watch out. The saṃprasāraṇa is in the abhyāsa only. The second half keeps dyut whole: didyute.
Rule. After dyut and the rest, parasmaipada optionally comes in luṅ. The plural dyudbhyaḥ shows that the whole group from dyut is meant. "Parasmaipada" comes from SK 2159.
Example.
dyut, luṅ, parasmaipada: aṅ by SK 2343 → adyutat "it shone"
dyut, luṅ, ātmanepada: sic → adyotiṣṭa.
Śvit "to be white": aśvitat or aśvetiṣṭa.
Applies when / blocked by. Aṅ comes only in the parasmaipada option. SK 2343 names parasmaipada endings, so in the ātmanepada only sic is possible. One of Bhaṭṭoji's commentators makes this point.
Watch out. These roots are ātmanepada roots. Outside luṅ, dyut keeps the ātmanepada: dyotate, didyute.
Rule. The i of mid takes guṇa before an affix that begins with an ś which is an it-letter. "Before a śit affix" comes from SK 2320.
Example.
mid "to be unctuous": mid + śap + te
→ śap begins with an it-letter ś, so i becomes e
→ medate.
Applies when / blocked by. Why "beginning with" an it-letter ś? Bhaṭṭoji answers with liṭ mimide. There the ending is eś (SK 2241). It has ś as it-letter, but not at its start. So this rule does not apply, and liṭ, being kit, has no guṇa of its own. One of Bhaṭṭoji's commentators adds: if any ś-marked affix counted, the substitutes of SK 2360 would wrongly come in pape as well.
luṅ: amidat or amediṣṭa, by SK 2345.
In the root list (roots 744–758):
rocate sūryaḥ "the sun shines"; haraye rocate bhaktiḥ "devotion pleases Hari". luṅ arucat or arociṣṭa.asrasat; ātmanepada asraṃsiṣṭa.vartate.Rule. After vṛt and the rest, parasmaipada optionally comes before sya and before san. "Parasmaipada" comes from SK 2159, and "optionally" from SK 2669.
Example.
vṛt, lṛṭ: with parasmaipada → vartsyati (see SK 2348)
→ without it → vartiṣyate.
Applies when / blocked by. The group has five roots, one of Bhaṭṭoji's commentators says. They are vṛt, vṛdh, śṛdh, syand and kṛp. The plural vṛdbhyaḥ shows that the group is meant.
Rule. After the four roots vṛt, vṛdh, śṛdh and syand, an ārdhadhātuka affix beginning with s takes no iṭ. This holds when there is no taṅ ending and no āna participle. "Iṭ" comes from SK 2184, "beginning with s" from SK 2506. The condition comes from the word "parasmaipada" in SK 2401, read as "no taṅ, no āna".
Example.
vṛt, lṛṭ, parasmaipada: vṛt + sya + ti
→ no iṭ; the ṛ takes guṇa (ar)
→ vartsyati "it will turn".
ātmanepada: taṅ is present, so iṭ comes → vartiṣyate.
lṛṅ: avartsyat, avartiṣyata. luṅ: avṛtat, avartiṣṭa.
Vṛdh and śṛdh go like vṛt, Bhaṭṭoji says.
Applies when / blocked by. Syand "to ooze" has ū as it-letter. That would make iṭ optional before s. Does this rule override the option? Bhaṭṭoji says yes, by the force of the word "four". If the option prevailed, naming four roots would be pointless. So with parasmaipada there is one form: syantsyati. The Bhāṣya does without the word "four", one of Bhaṭṭoji's commentators notes; scholars discuss this.
Where taṅ is present, the rule stands aside, and the ū-option works:
syandiṣyate or syantsyate;syandiṣīṣṭa or syantsīṣṭa;asyandiṣṭa or asyanta, dual asyantsātām, plural asyantsata;asyandiṣyata or asyantsyata.In luṅ with parasmaipada (SK 2345), aṅ comes and the n is lost: asyadat. Liṭ has both too: sasyandiṣe or sasyantse.
Watch out. The condition is "no taṅ and no āna", not simply "parasmaipada". One of Bhaṭṭoji's commentators explains why. The rule must also reach forms whose ending has been deleted, such as the desiderative imperative vivṛtsa (hi lost by luk).
Rule. After anu, vi, pari, abhi and ni, the s of syand is optionally replaced by ṣ. This holds when the agent is not a living being. "The s" comes from SK 335, and "a retroflex" from the heading SK 210.
Example.
anu + syandate jalam "the water flows along"
→ the agent, water, is not a living being
→ anuṣyandate jalam or anusyandate jalam.
Applies when / blocked by. Why "not living beings"? Bhaṭṭoji answers: anusyandate hastī "the elephant moves along". The agent lives, so there is no ṣ.
What if the agents are mixed? Bhaṭṭoji gives matsyodake anuṣyandete "the fish and the water flow along". The ṣ is still allowed as an option. The reason is the kind of negation in aprāṇiṣu. It is a paryudāsa: "for what has an agent other than a living being". Such a pair still has a non-living agent. Had the rule said "not in living beings" as an outright prohibition (prasajyapratiṣedha), there would be no ṣ here.
Watch out. A paryudāsa picks out "something other than X". A prasajyapratiṣedha forbids the operation wherever X is present. The two can give different results, as here.
Rule. The rule is split as kṛpa u raḥ laḥ, and kṛpa is taken twice. This gives two statements.
Example.
kṛp "to be able", laṭ: kṛp + śap + te
→ guṇa: karp + a + te
→ the r becomes l
→ kalpate.
liṭ: cakṛp + e
→ liṭ is kit, so there is no guṇa; the ṛ itself becomes ḷ
→ cakḷpe.
Second person singular cakḷpiṣe or cakḷpse, with and without iṭ, "as for syand".
Applies when / blocked by. Why the second statement? One of Bhaṭṭoji's commentators explains. With the first alone, cakḷpe could not be derived: there is no r in cakṛpe, only the vowel ṛ. So the words are repeated, and the second reading reaches the ṛ.
This rule's "r … l" was borrowed by SK 2326 (palāyate).
Watch out. From here on the root is often named kḷp, as in the next two rules. It is the same root kṛp.
Rule. After kḷp, parasmaipada optionally comes in luṭ. By "ca", it also comes before sya and san, drawn in from SK 2347. "Parasmaipada" comes from SK 2159, and "optionally" from SK 2669.
Example.
kḷp, luṭ: parasmaipada kalptāsi "you will be able" (see SK 2352)
→ ātmanepada kalpitāse or kalptāse.
Rule. After kḷp, the affix tās takes no iṭ. By "ca", neither does an ārdhadhātuka affix beginning with s. This holds when there is no taṅ and no āna. The words come from SK 2184, SK 2506, SK 2348 and SK 2401.
Example.
luṭ, parasmaipada: kḷp + tās + si
→ no iṭ; guṇa: karp
→ r → l by SK 2350: kalp
→ kalptāsi; plural kalptāstha.
luṭ, ātmanepada: taṅ is present, so the ū-option works: kalpitāse or kalptāse.
Applies when / blocked by. Bhaṭṭoji gives every lakāra.
kalpsyati; ātmanepada kalpiṣyate or kalpsyate.kalpiṣīṣṭa or kḷpsīṣṭa. The second has no guṇa, its affix being kit.akḷpat, with aṅ (SK 2345); ātmanepada akalpiṣṭa or akḷpta.akalpsyat; ātmanepada akalpiṣyata or akalpsyata.After this rule the list says "vṛt": "finished". The dyutādi and the vṛtādi end here.
Then begin the ghaṭādi, from ghaṭ "to strive". The first thirteen, up to tvar, are ātmanepada and ṣit (ṣ-marked). All the ghaṭādi are mit, by a later list-rule, ghaṭādayo mitaḥ. A mit root shortens its vowel before ṇi (SK 2568). With ciṇ and ṇamul it may lengthen it again (SK 2762). So ghaṭ gives ghaṭayati and vighaṭayati, with short a.
What of udghāṭana, with long ā? Bhaṭṭoji says it comes from a different root, the tenth-class ghaṭ "to join". That root is not mit, by the list-rule nānye mito 'hetau.
Watch out. The condition is "no taṅ, no āna", as in SK 2348. It is not simply "when parasmaipada is used".
vyath → vivyathe; phaṇ → pheṇatuḥ / paphaṇatuḥ; ghaṭ → ghaṭayatiThis group finishes the ghaṭādi (roots 763–821) and adds the phaṇādi, seven roots from phaṇ. Vyath "to fear" keeps a vowel in its abhyāsa: vivyathe. The ghaṭādi are all mit, so their vowel stays short before ṇi: ghaṭayati. A set of list-rules (gaṇasūtras) adds other mit roots and takes some away. Last, the seven phaṇādi may take e and drop the abhyāsa in liṭ: pheṇatuḥ or paphaṇatuḥ.
Rule. In liṭ, the abhyāsa of vyath takes saṃprasāraṇa. "Saṃprasāraṇa" comes from SK 2344, and "of the abhyāsa" from SK 2620.
Example.
vyath vyath e → in the abhyāsa, y becomes i: viath
→ pūrvarūpa: vi, and halādiḥ śeṣaḥ removes the th
→ vi vyath e
→ vivyathe "he feared".
Applies when / blocked by. This rule is an exception to halādiḥ śeṣaḥ (SK 2179), Bhaṭṭoji says. But it blocks only the loss of the y. The th is still removed by halādiḥ śeṣaḥ. One of Bhaṭṭoji's commentators adds that the v takes no saṃprasāraṇa: no saṃprasāraṇa is made before another one (SK 363). Scholars discuss this point.
Why "in liṭ"? One of Bhaṭṭoji's commentators gives the desiderative vivyathiṣate, where the i comes by SK 2317 instead.
In the root list, Bhaṭṭoji now gives the rest of the ghaṭādi. Many are roots from other classes, restated here in a particular sense to make them mit.
akṣañji or akṣāñji.hiḍayati; in the sense "disregard", heḍayati.darayati "he frightens"; elsewhere dārayati.śrapayati; elsewhere śrāpayati.madayati; elsewhere mādayati.The list-rules that close the group:
ghaṭā, vyathā.jvalayati or jvālayati; but with a prefix only prajvalayati.kāmayate, āmayati, cāmayati.niśāmayati rūpam "he sees the form".āyāmayati "he lengthens", but yamayati brāhmaṇān "he feeds the brāhmaṇas".avaskhādayati.Svāmin construes the last three rules differently. Bhaṭṭoji rejects his view, as against the Vṛtti and the Nyāsa.
Watch out. A mit root keeps its vowel short before ṇi where one would expect vṛddhi or lengthening. Always check whether a root is mit before forming its causative.
Rule. The seven roots from phaṇ optionally take e for their a and lose the abhyāsa. This happens before a kit liṭ ending and before thal with iṭ. The seven are phaṇ, rāj, bhrāj, bhrāś, bhlāś, syam and svan. "e" comes from SK 2471, "kit" from SK 2363, and "optionally" from SK 2356.
Example.
phaṇ "to go", liṭ dual: phaṇ phaṇ atus
→ with the option: e for a, and the abhyāsa lost
→ pheṇatuḥ
→ without it: paphaṇatuḥ.
Plural pheṇuḥ or paphaṇuḥ. Second person singular with iṭ: pheṇitha.
Phaṇ is mit, so its causative has short a: phaṇayati.
Applies when / blocked by. The other six follow.
rejatuḥ or rarājatuḥ; reje or rarāje. Here even a long ā takes e. The word "short a" is carried forward, yet the rule would be pointless for rāj otherwise. So it works by the force of being taught (vidhānasāmarthya).bhreje or babhrāje.bhrāśyate or bhrāśate (SK 2321); liṭ bhreśe or babhrāśe.bhleśe or babhlāśe.syematuḥ or sasyamatuḥ.svenatuḥ or sasvanatuḥ.The phaṇādi end with svan. So dhvan, next in the list, has only dadhvanatuḥ.
Here the list says "vṛt" again: the ghaṭādi are complete. Some place this "vṛt" before phaṇ. On that view phaṇ is not mit. Which causative that view gives, scholars discuss.
Then begin the jvalādi. Jval "to shine" gives ajvālīt, with vṛddhi by SK 2330. Patḷ "to fall" gives papāta, petatuḥ.
Watch out. Why is phaṇ read after the list-rule na kamyamicamām? Bhaṭṭoji says the "not" has lapsed there, since nothing can be forbidden before it applies. His commentators add that only "mit" continues. Phaṇ was placed there for the sake of the phaṇādi operations of this rule.
sah + tā → soḍhā → parisoḍhāThis group takes three roots of the first class. Pat "to fall" takes an augment in its aṅ-aorist: apaptat. Bhram "to wander" may take e and drop the abhyāsa in liṭ: bhrematuḥ. Sah "to bear" loses its final in luṭ, and its a becomes o: soḍhā. After pari its s does not become ṣ: parisoḍhā. But in other forms it may: paryaṣahata. The running example is sah: sah + tā → soḍhā, and with pari, parisoḍhā.
Rule. When aṅ follows, the augment pum is added to pat. "When aṅ follows" comes from SK 2406. The m of pum is an it-letter, and its u is for pronunciation. An m-marked augment goes after the last vowel (SK 37).
Example.
patḷ, luṅ: the root has ḷ as it-letter, so aṅ replaces cli (SK 2343): a + pat + a + t
→ pum goes after the vowel of pat: a + pa p t + a + t
→ apaptat "he fell".
With pra and ni: praṇyapaptat, with ṇ by SK 2285.
Applies when / blocked by. In the root list that follows:
akvathīt. So also path (apathīt) and math (amathīt).methatuḥ.vavāma, vavamatuḥ. The Bhāgavṛtti also gives vematuḥ, but Bhaṭṭoji says it is not seen in the Bhāṣya.bhramyati or bhramati.Watch out. Pum is an augment, not a substitute. Nothing of pat is lost; a p is inserted after its vowel.
Rule. For jṝ, bhram and tras, two things happen optionally: e replaces the a, and the abhyāsa is lost. This holds before a kit liṭ ending and before thal with iṭ. "e" comes from SK 2471, "in liṭ" from SK 2260, "kit" from SK 2363, and "thal with iṭ" from SK 2261.
Example.
bhram, liṭ dual: bhram bhram atus
→ with the option: bhrematuḥ
→ without it: babhramatuḥ.
luṅ: abhramīt, with no vṛddhi by SK 2299.
Applies when / blocked by. One of Bhaṭṭoji's commentators calls this an option where the operation was not otherwise due.
In the root list that follows, kṣar "to move" gives akṣārīt, with vṛddhi by SK 2330. Then come two ātmanepada roots. The first is ṣah "to bear".
pariṣahate.sehe; luṭ with iṭ sahitā.sah + tā → h becomes ḍh (SK 324): saḍh + tā
→ t becomes dh (SK 2280): saḍh + dhā
→ ṣṭutva (SK 113): saḍh + ḍhā
→ the first ḍh is lost (SK 2335): sa + ḍhā.
The next rule finishes the form.
Rule. When a ḍh has been deleted, the a of sah and of vah is replaced by o. "When ḍh has been deleted" comes from SK 174. That rule also mentions deletion of r, but that part is not carried forward, since it cannot apply here.
Example.
sa + ḍhā (from SK 2356) → the ḍh after sah has been deleted
→ the a becomes o
→ soḍhā "he will bear".
Vah "to carry" gives, in the same way, voḍhā and uvoḍha (SK 2409).
Applies when / blocked by. Without the deletion of ḍh there is no o. One of Bhaṭṭoji's commentators gives sahate and vahate.
Watch out. In other roots the vowel before a deleted ḍh is lengthened (SK 174). With sah and vah, o comes instead.
Rule. The s of sah is not replaced by ṣ when the root has the form soḍh. "Not" is carried forward from SK 3168.
Example.
pari + soḍhā → SK 2275 would make the s after pari into ṣ
→ this rule forbids it, since the root has the form soḍh
→ parisoḍhā "he will endure".
Applies when / blocked by. Compare pariṣahate (SK 2356). There the root has the form sah, so the ṣ comes.
Rule. After pari, ni and vi, the s of siv and the rest is optionally replaced by ṣ. This holds even when the augment aṭ stands between. "The rest" are sah, suṭ, stu and svañj. The prefixes are carried forward from SK 2275.
Example.
sah with pari, laṅ: pari + a + sahata
→ aṭ stands between pari and the s
→ paryasahata or paryaṣahata "he endured".
Applies when / blocked by. One of Bhaṭṭoji's commentators gives more: paryaṣīvyat or paryasīvyat, nyaṣṭaut or nyastaut, vyaṣvajat or vyasvajat.
Next comes ram "to play", an ātmanepada root. liṭ reme, remiṣe; luṭ rantā; lṛṭ raṃsyate; āśīrliṅ raṃsīṣṭa; luṅ araṃsta. In remiṣe the iṭ comes by the krādi restriction (SK 2293). Then begin the parasmaipada roots up to kas, starting with ṣad "to sink".
Watch out. The option is what this rule adds. Where no aṭ stands between, Bhaṭṭoji gives only pariṣahate (SK 2356).
sad + śap + ti → sīdati; ni + sīdati → niṣīdati; liṭ niṣasādaEleven roots take a new stem before a śit affix. Pā "to drink" becomes piba, sthā "to stand" becomes tiṣṭha, sad "to sit" becomes sīda, and so on. These stems give the familiar present forms. Then two rules finish sad and śad. After a prefix, the s of sad following the abhyāsa does not become ṣ: niṣasāda. And śad "to wither" takes ātmanepada only where a śit affix follows: śīyate, but śaśāda. The running example is sad: sad + śap + ti → sīdati; with ni, niṣīdati; liṭ niṣasāda.
Rule. Before an affix beginning with an ś that is an it-letter, eleven roots are replaced by eleven stems, in order:
"Before a śit affix" comes from SK 2320.
Example.
sad + śap + ti → śap begins with an it-letter ś, so sad is replaced by sīda
→ sīda + a + ti
→ sīdati "he sits".
Applies when / blocked by. The substitutes come only before a śit affix. Elsewhere sad stays:
sasāda, sedatuḥ, second person singular seditha or sasattha;sattā; lṛṭ satsyati;asadat, with aṅ, since sad has ḷ as it-letter.After ni, the s becomes ṣ by SK 2271: niṣīdati. In laṅ it does so across the augment, by SK 2276: nyaṣīdat.
Bhaṭṭoji's commentators read śit as "an ś that is an it-letter", at the start of the affix. Why not "an affix with ś as it-letter" anywhere? Then piba would come in pape, the passive liṭ of pā. Its ending eś has ś as it-letter at the end.
The other stems return later in this chapter: pibati, jighrati, dhamati, tiṣṭhati, manati, praṇiyacchati (SK 2378); dhāvati and ṛcchati (SK 2382); paśyati (SK 2403); śīyate (SK 2362). Gam uses another rule: gacchati (SK 2400).
Watch out. The stems replace the whole root. Piba ends in a, so pibati has no guṇa.
Rule. In liṭ, the s of sad that follows the abhyāsa is not replaced by ṣ.
Example.
ni + sasāda → the s of the abhyāsa takes ṣ by SK 2277: ni ṣa sāda
→ the s after the abhyāsa is protected by this rule
→ niṣasāda "he sat down".
Dual niṣedatuḥ: the abhyāsa is lost, so there is no s after an abhyāsa.
Applies when / blocked by. Why "in liṭ"? One of Bhaṭṭoji's commentators gives the desiderative niṣiṣatsati, where the ṣ does come.
Some read the rule as sadisvañjoḥ, naming svañj too. One of Bhaṭṭoji's commentators rejects that reading; svañj is brought in later by a vārttika (SK 2397). Scholars discuss this.
Next comes the root śad. Its dhātupāṭha gloss is "śātana". Bhaṭṭoji says the root means falling apart, and śātana names its domain. One of his commentators explains that śātana is a familiar form of the root itself.
Watch out. Only the second s is protected. The first, in the abhyāsa, still becomes ṣ.
Rule. After śad, ātmanepada comes when a śit affix is to follow it. "Ātmanepada" comes from SK 2157.
Example.
śad + śap + te → a śit affix, śap, follows, so the ending is ātmanepada
→ śad is replaced by śīya (SK 2360)
→ śīyate "it withers".
Applies when / blocked by. Bhaṭṭoji says "when a śit affix is to follow" (śidbhāvin). One of Bhaṭṭoji's commentators explains why. The ending is chosen before śap is added, and śap depends on that ending. So the rule must look ahead to the śit affix.
Where no śit affix will come, the ending is parasmaipada. liṭ śaśāda, śedatuḥ, śeditha or śaśattha; luṭ śattā; luṅ aśadat.
In the root list that follows:
krośati, kroṣṭā, luṅ akrukṣat with ksa.roḍhā, rokṣyati, arukṣat.akāsīt or akasīt. Here the jvalādi end.Then the roots with a svarita it-vowel, up to guh, which take both padas. Budhir "to know" has ir as it-marker, so aṅ is optional: abudhat or abodhīt, and abodhiṣṭa. It does not take the ciṇ of SK 2328.
Watch out. Śad is not an ātmanepada root in general. Its ātmanepada is tied to the śit affix.
guh + śap + ti → gūhati; aguh + ksa + ta → aghukṣata / agūḍhaFive roots lose their penultimate a before a vowel-initial kit or ṅit affix: cakhnatuḥ, jagmatuḥ. Guh "to hide" lengthens its u to ū where guṇa would come: gūhati. Its aorist without iṭ takes ksa: aghukṣata. For four roots, guh among them, that ksa may disappear before a dental ending: agūḍha. The running example is guh: guh + śap + ti → gūhati, and luṅ aghukṣata or agūḍha.
Rule. The penultimate of gam, han, jan, khan and ghas is deleted before a kit or ṅit affix beginning with a vowel. It is not deleted before aṅ. "Penultimate" comes from SK 2364, and "beginning with a vowel" from SK 271.
Example.
khan "to dig", liṭ dual: ca khan atus
→ atus is kit and begins with a vowel, so the penultimate a of khan is lost
→ cakhnatuḥ.
One of Bhaṭṭoji's commentators adds jagmatuḥ, jaghnatuḥ, jajñe, jakṣatuḥ.
Applies when / blocked by. "kit or ṅit": ṇal is neither, so cakhāna keeps its a. "Not before aṅ": agamat keeps its a.
Before y, the vowel-initial condition fails. There khan follows SK 2319 instead: āśīrliṅ khāyāt or khanyāt.
In the root list (roots 879–896), vyay gives avyayīt, with no vṛddhi (SK 2299). Laṣ gives laṣyati or laṣati (SK 2321). Chaṣ gives cacchaṣatuḥ, keeping its abhyāsa. The list ends with guh "to hide".
Watch out. Only the penultimate goes, not the whole syllable: khan becomes khn.
Rule. The penultimate u of guh is replaced by ū before a vowel-initial affix that causes guṇa. The rule names the root as goh, its guṇa form. That shows it applies only where guṇa is due. "Beginning with a vowel" comes from SK 271.
Example.
guh + śap + ti → śap begins with a vowel and would cause guṇa
→ instead of guṇa, the u becomes ū
→ gūhati "he hides"; ātmanepada gūhate.
Applies when / blocked by. The rule is an exception to guṇa, one of Bhaṭṭoji's commentators says.
Guh has ū as it-letter, so iṭ is optional. That changes the forms.
gūhitā. The affix now begins with a vowel, so ū comes.goḍhā. No vowel-initial affix, so plain guṇa.gūhiṣyati or ghokṣyati.gūhet.guhyāt: the affix begins with y and is kit, so neither ū nor guṇa.agūhīt; without iṭ, ksa comes (SK 2336): aghukṣat.Watch out. Where no guṇa is due, there is no ū. In liṭ dual the ending is kit: juguhatuḥ, with short u.
Rule. For duh, dih, lih and guh, the affix ksa is optionally deleted by luk before an ātmanepada ending beginning with a dental. The whole ksa goes.
Example.
guh, luṅ ātmanepada: aguh + ksa + ta
→ ta begins with a dental, so ksa may be deleted: aguh + ta
→ h becomes ḍh, t becomes dh, ṣṭutva makes it ḍh, the first ḍh is lost, and the u is lengthened (SK 174)
→ agūḍha.
Without the deletion: aghukṣata.
Applies when / blocked by. Before ātām, which does not begin with a dental, ksa stays. Its a is lost by SK 2337: aghukṣātām. Plural aghukṣanta.
First person dual: aguhvahi with deletion, the ending counting as dental-initial, or aghukṣāvahi. First person plural: only aghukṣāmahi, since m is not a dental.
One of Bhaṭṭoji's commentators gives the other roots: adugdha or adhukṣata, adigdha or adhikṣata, alīḍha or alikṣata.
Then come the vowel-final roots that take both padas.
śrayati, śrayate, śiśriyatuḥ, śrayitā, and luṅ aśiśriyat, with caṅ by SK 2312.bharati; liṭ babhāra, babhratuḥ, babhartha, babhṛva; ātmanepada babhṛṣe; luṭ bhartā.Watch out. In aguhvahi there is no ū. No vowel-initial affix follows the root, so SK 2364 cannot apply.
bhṛ → bhariṣyati, bhriyāt, bhṛṣīṣṭa, abhṛtaThis group follows bhṛ "to bear" through the lakāras where an ṛ-final root needs special rules. Bhṛ is aniṭ, yet in lṛṭ its sya takes iṭ: bhariṣyati. In āśīrliṅ its ṛ becomes ri: bhriyāt. The ātmanepada āśīrliṅ and luṅ come next: bhṛṣīṣṭa, abhṛta.
Rule. After a root ending in ṛ, and after han, the affix sya takes iṭ. "Iṭ" comes from SK 2184.
Example.
bhṛ, lṛṭ: bhṛ + sya + ti
→ sya takes iṭ: bhṛ + i + sya + ti
→ guṇa of ṛ, and s becomes ṣ after i
→ bhariṣyati "he will bear".
Applies when / blocked by. Bhṛ is an aniṭ root, so SK 2246 would forbid iṭ. One of Bhaṭṭoji's commentators calls this rule an exception to that prohibition. He adds ātmanepada bhariṣyate.
The same rule later gives hvariṣyati (SK 2379) and svariṣyati (SK 2381).
Watch out. The rule names sya only. In luṭ bhṛ stays aniṭ: bhartā.
Rule. A final ṛ of an aṅga is replaced by riṅ, heard as ri. This happens before three things. They are the affix śa, the affix yak, and a liṅ that begins with y and is ārdhadhātuka. "ṛ" comes from SK 1234. "Beginning with y" (SK 2649) and "ārdhadhātuka" (SK 2298) qualify the liṅ only.
Example.
bhṛ, āśīrliṅ: bhṛ + yāt
→ ṛ is replaced by ri
→ bhriyāt "may he bear".
Applies when / blocked by. Why is the i of ri not lengthened by SK 2298? Bhaṭṭoji answers. The long rī (rīṅ, SK 1234) was already available. Teaching the short ri would be pointless if it were then lengthened. So it stays short.
One of Bhaṭṭoji's commentators gives the other two cases. With śa: mriyate. With yak: kriyate, mriyate. The words "beginning with y" and "ārdhadhātuka" cannot go with śa, and are pointless with yak, so they qualify liṅ alone.
The same rule gives sriyāt and griyāt later (SK 2381, SK 2384).
Watch out. Vidhiliṅ is sārvadhātuka, so it does not count. Its form keeps śap and guṇa: bharet.
bhṛ → bhariṣyati, bhriyāt, bhṛṣīṣṭa, abhṛta (continued)The root bhṛ "to bear" (bhṛñ) is still under study. Its lṛṭ bhariṣyati and its āśīrliṅ bhriyāt are already derived. Two rules remain. SK 2368 keeps guṇa out of the ātmanepada āśīrliṅ bhṛṣīṣṭa and the ātmanepada luṅ. SK 2369 then removes the s of the luṅ to give abhṛta "he bore (for himself)".
Rule. After a root in ṛ or ṝ, two affixes count as kit, that is, as if marked with k. They are a liṅ affix, and a sic followed by an ātmanepada ending (a taṅ ending). Both must begin with a jhal, a consonant that is neither a nasal nor a semivowel.
The words "liṅ and sic before ātmanepada" are carried forward from SK 2300. The word jhal is carried forward from SK 2612. Bhaṭṭoji's commentators both say so. A kit affix blocks guṇa of the root vowel.
Example. The āśīrliṅ ātmanepada of bhṛ:
bhṛ + sīṣṭa → sīṣṭa begins with the jhal s, and follows ṛ: it is kit
→ no guṇa of the ṛ
→ bhṛṣīṣṭa "may he bear (for himself)". Likewise bhṛṣīyāstām (dual).
The parasmaipada luṅ is not covered. There sic is not followed by a taṅ ending, so it is not kit. Vṛddhi comes by SK 2297, and r follows the vowel:
a + bhṛ + s + ī + t → vṛddhi → abhārṣīt "he bore". Likewise abhārṣṭām, abhārṣuḥ.
Applies when / blocked by. The word "beginning with a jhal" governs both affixes. So a sic that has taken the augment iṭ begins with i, not with a jhal. It is not kit.
Watch out. The rule looks at the ātmanepada only. Do not strip the vṛddhi from abhārṣīt: that is a parasmaipada form.
Rule. After a stem (aṅga) that ends in a short vowel, sic is deleted when a jhal follows. The word sic is supplied following the Bhāṣya, Patañjali's great commentary. "Deletion" (lopa) is carried forward from SK 54, and "before a jhal" from SK 2281.
Example.
a + bhṛ + s + ta → sic is kit by SK 2368, so no guṇa
→ the stem abhṛ ends in short ṛ, and the jhal t follows the s
→ the s is deleted → abhṛta "he bore (for himself)".
In the dual, a + bhṛ + s + ātām → ātām begins with a vowel, so the s stays → abhṛṣātām.
The lṛṅ: abhariṣyat.
Applies when / blocked by. Why "a short vowel"? Bhaṭṭoji's commentators both give acyoṣṭa, where the stem vowel is not short and sic stays. The deletion falls on the affix sic alone. One of Bhaṭṭoji's commentators says this is why the word aṅga "stem" is in the rule; scholars discuss this.
Bhaṭṭoji then gives three more roots of this kind.
jahartha, jahriva, jahriṣe, hartā, hariṣyati.dharati, adhārṣīt, and the ātmanepada luṅ adhṛta, made like abhṛta.ninayitha or ninetha, with and without iṭ; and ninyiṣe.Then Bhaṭṭoji announces a new run of roots: the vowel-final roots that take the parasmaipada. The first is dheṭ "to drink": dhayati "he drinks".
dhe → dhā → dadhau, dadhatuḥ, dheyāt, adadhat / adhāt / adhāsītMany roots are taught ending in e or ai, yet behave as ā-final roots. This group shows how. SK 2370 turns the final into ā. SK 2371 and SK 2372 then give the liṭ forms dadhau, dadhatuḥ. SK 2373 names the roots dā and dhā "ghu". SK 2374 gives the āśīrliṅ dheyāt. SK 2375 to SK 2377 give the three luṅ forms of dheṭ: adadhat, adhāt, adhāsīt. SK 2378 makes e optional in glai: glāyāt or gleyāt.
Rule. A root that ends in ec (e, o, ai or au) in its taught form (upadeśa) takes ā in place of that vowel. This does not happen when a śit affix follows, an affix with the marker ś.
The word "root" (dhātoḥ) is carried forward from SK 2177. Bhaṭṭoji's commentators both say so.
Example.
dhe "to drink" (taught as dheṭ) → ends in e in its taught form
→ dhā; the ā is withheld only before a śit.
But dhe + śap + ti → śap is śit, so the e stays → dhayati "he drinks".
In the same way glai "to be weary" → glā.
Applies when / blocked by. Why "in its taught form"? In cetā "gatherer" and stotā "praiser" the roots are ci and stu. Their e and o come from guṇa, so they take no ā. Why "a root"? The noun go "cow" gives gobhyām, with no ā. Bhaṭṭoji's commentators give these counter-examples.
Watch out. Bhaṭṭoji's gloss is "but not when a śit follows". His commentators read it this way: the ā needs no following affix at all. It is simply withheld before a śit.
Rule. After a stem ending in ā, the liṭ ending ṇal is replaced by au. ṇal is the prathama singular ending of liṭ.
Example.
dhā + ṇal → doubling, and the rules of the abhyāsa → dadhā + a
→ au for ṇal → dadhā + au
→ vṛddhi of ā and au → dadhau "he drank".
So also jaglau "he was weary", from glai.
Applies when / blocked by. One of Bhaṭṭoji's commentators says the rule applies at the stage dadhā + a, after the doubling.
Rule. The ā of a stem is deleted before a vowel-initial ārdhadhātuka affix that is kit or ṅit. It is also deleted before iṭ. One of Bhaṭṭoji's commentators takes this iṭ to include both the ending iṭ and the augment iṭ.
"Ārdhadhātuka" is a heading here. "Before a vowel" and "kit or ṅit" are carried forward from SK 2507. The kit-or-ṅit condition does not qualify iṭ, which is named separately.
Example.
dhā + atus → doubling first → dadhā + atus
→ atus is vowel-initial and kit, so the ā is deleted
→ dadhatuḥ "they two drank". Likewise dadhuḥ.
Before thal with iṭ: dadhā + i + tha → dadhitha. Without iṭ: dadhātha.
Before va with iṭ: dadhiva "we two drank".
The luṭ: dhātā "he will drink".
Applies when / blocked by. This deletion is the later rule. So it would apply before doubling. Then dh + atus would have no vowel left to double. SK 2243, dvirvacane 'ci, prevents this. Doubling is done first, and only then is the ā deleted.
Watch out. In dadhau the ending ṇal is neither kit nor ṅit. So the ā stays, and SK 2371 can give au. Had the ā gone, there would be no au.
Rule. This is a saṃjñā, a naming rule. Roots having the form dā and roots having the form dhā are called ghu. The roots dāp "to cut" and daip "to cleanse" are excluded.
Example. Bhaṭṭoji's commentators list the ghu roots.
Of the form dā: ḍudāñ "to give", dāṇ "to give", do "to cut" and deṅ "to protect".
Of the form dhā: ḍudhāñ "to hold, to nourish" and dheṭ "to drink".
So dhe → dhā (SK 2370) → dheṭ is ghu.
Applies when / blocked by. do, deṅ and dheṭ have the form dā or dhā only after SK 2370 gives ā. They still count. daip is excluded, so Bhaṭṭoji later gives dāyāt and adāsīt for it, without the ghu rules.
Watch out. The name matters at once. SK 2374, SK 2375, SK 2376 and SK 2389 all name the ghu roots or dheṭ.
Rule. The ghu roots, and mā, sthā, gā, pā, hā and sā, take e for their final. This happens before a liṅ affix that is ārdhadhātuka and kit, as in the āśīrliṅ.
The list "ghu, mā, sthā …" is carried forward from SK 2462. "Kit" is carried forward from SK 2507. Only "kit" is carried, not "ṅit": an ārdhadhātuka liṅ cannot be ṅit.
Example.
dhā + yāt → āśīrliṅ, ārdhadhātuka and kit
→ e for the final ā → dheyāt "may he drink". Likewise dheyāstām, dheyāsuḥ.
One of Bhaṭṭoji's commentators adds deyāt, stheyāt, geyāt, peyāt, heyāt.
Applies when / blocked by. Why "ārdhadhātuka"? In the vidhiliṅ of mā the affix is sārvadhātuka: māyāt. Why "kit"? The ātmanepada āśīrliṅ dāsīṣṭa is not kit, so the ā stays. One of Bhaṭṭoji's commentators gives both.
Rule. After dheṭ and śvi "to grow", caṅ optionally replaces cli in a luṅ that denotes the agent. cli is the placeholder affix that a root takes before luṅ (SK 2221). caṅ is its reduplicating substitute (SK 2312).
Example.
dhā + luṅ → caṅ for cli → a + dhā + a + t
→ doubling by SK 2315, caṅi → a + dadhā + a + t
→ ā deleted by SK 2372 → adadhat "he drank". Likewise adadhatām.
For śvi, one of Bhaṭṭoji's commentators gives aśiśviyat (see SK 2421).
Applies when / blocked by. The rule is optional. In the other option, cli is not replaced by caṅ. SK 2376 and SK 2377 then give the forms of that option.
Rule. After ghrā "to smell", dheṭ, śā "to make thin", chā "to cut" and sā "to end", sic is optionally deleted by luk before a parasmaipada ending. Luk is deletion of an affix (SK 2202). "Luk" is carried forward from SK 1276. "Of sic, before parasmaipada" comes from SK 2223.
Example. The option in which caṅ does not come:
a + dhā + s + t → luk of sic → adhāt "he drank". Likewise adhātām.
In the plural jus replaces jhi (SK 2227): a + dhā + us → the ā merges into the u (SK 2214) → adhuḥ.
So too aghrāt "he smelt".
Applies when / blocked by. For dheṭ, SK 2223 had already made this luk obligatory. For the others it was not due at all. One of Bhaṭṭoji's commentators says the rule serves both cases. Why "before parasmaipada"? The ātmanepada form vyatyaghrāsta keeps its s.
Watch out. śā, chā and sā are named here in their ā form. They are taught as śo, cho and ṣo.
Rule. The roots yam, ram, nam and the ā-final roots take the augment sak. Also, the sic after them takes iṭ. Both happen before a parasmaipada ending. The k of sak is a marker, so the s is added at the end of the root. Bhaṭṭoji's commentators name two different sources for the carried-forward iṭ; scholars discuss this.
Example. The option in which sic is not deleted by SK 2376:
a + dhā + s + t → sak to the root, iṭ to sic, and īṭ before t
→ a + dhā + s + i + s + ī + t
→ the s of sic is lost between iṭ and īṭ (SK 2266) → adhāsīt "he drank".
Likewise adhāsiṣṭām, adhāsiṣuḥ.
One of Bhaṭṭoji's commentators adds ayaṃsīt from yam and anaṃsīt from nam.
Applies when / blocked by. Only before parasmaipada. The ātmanepada luṅ of ram is araṃsta, without sak or iṭ.
Bhaṭṭoji then gives glai and mlai "to be weary". Their sense is loss of strength. The ā of SK 2370 does not reach glāyati, since śap is śit. In liṭ, jaglau has au by SK 2371. In the madhyama singular, jaglitha or jaglātha.
Rule. A root in ā that begins with a consonant cluster (saṃyoga) takes e optionally, before a kit ārdhadhātuka liṅ. It applies only to roots other than ghu, mā, sthā and the rest of SK 2462. For those, SK 2374 makes the e obligatory.
Example.
glā + yāt → glā begins with the cluster gl, and is not in the SK 2462 list
→ glāyāt or gleyāt "may he be weary". So too ghrāyāt or ghreyāt.
The luṅ has sak and iṭ by SK 2377: aglāsīt.
Applies when / blocked by. Why "other"? sthā begins with a cluster too, but it is in the SK 2462 list. So only stheyāt is correct. Why "beginning with a cluster"? rai "to sound" has none, so only rāyāt. Bhaṭṭoji's commentators both give stheyāt; one of them gives rāyāt.
Bhaṭṭoji now lists the ai-final roots and the true ā-final roots. Three points stand out.
sāyāt, asāsīt.pāyāt, apāsīt.dāyāt, adāsīt.The roots pā, ghrā, dhmā, sthā, mnā and dāṇ take their special present stems before a śit affix, by SK 2360. So pibati "he drinks", jighrati, dhamati, tiṣṭhati, manati and praṇiyacchati. piba ends in a, so it takes no guṇa of the penultimate vowel. Outside the śit domain the root returns: peyāt, apāt, adhitaṣṭhau, stheyāt, deyāt, adāt.
Watch out. styai and ṣṭyai both give styāyati. The difference shows in forms like tiṣṭyāsati and atiṣṭyapat.
hvṛ → jahvāra, jahvaratuḥ, hvaryāt, ahvārṣīt; ṛ → āra, āritha, aryāt, ārṣītThis group works through roots ending in ṛ, and a few ending in u. Two model roots carry it: hvṛ "to be crooked" and ṛ "to go". SK 2379 and SK 2383 give guṇa in liṭ, even before kit endings: jahvaratuḥ, āratuḥ. SK 2380 gives guṇa before the y of āśīrliṅ: hvaryāt, aryāt. SK 2381, SK 2384 and SK 2385 settle when iṭ comes: in thal (āritha) and in sic (asāvīt). SK 2382 decides which roots take aṅ in luṅ.
Rule. A stem that begins with a consonant cluster and ends in ṛ takes guṇa before liṭ. The word "liṭ" is carried forward from SK 2388.
Example.
hvṛ + atus → doubling → jahvṛ + atus → atus is kit, but this rule still gives guṇa
→ ar for ṛ → jahvaratuḥ "they two were crooked". Likewise jahvaruḥ, jahvartha.
hvṛ + ṇal → jahvṛ + a → guṇa → ar, with r following (raparatva)
→ vṛddhi of the penultimate a (SK 2282) → jahvāra "he was crooked".
Outside liṭ: hvartā (luṭ), and hvariṣyati (lṛṭ) with iṭ by SK 2366.
Applies when / blocked by. The rule is meant for kit endings like atus. Before ṇal, SK 254 (aco ñṇiti) would give vṛddhi of the final. But this guṇa is the later rule, so it wins before ṇal too. Then the penultimate a takes vṛddhi, and jahvāra results. Why "beginning with a cluster"? kṛ has no cluster, so its liṭ is cakratuḥ, without guṇa.
Rule. Guṇa replaces the final vowel of the root ṛ (arti), and of a cluster-initial stem ending in ṛ. This happens before yak, and before a y-initial ārdhadhātuka liṅ, as in āśīrliṅ. It sets aside the riṅ of SK 2367.
Example.
hvṛ + yāt → yāt is kit, so guṇa would be barred, and SK 2367 would give riṅ
→ this rule gives guṇa instead → hvaryāt "may he be crooked".
The parasmaipada luṅ has vṛddhi by SK 2297: ahvārṣīt, ahvārṣṭām.
Applies when / blocked by. Why "beginning with a cluster"? kṛ gives kriyāt. Why "y-initial"? saṃskṛṣīṣṭa has no guṇa. One of Bhaṭṭoji's commentators gives these.
Bhaṭṭoji next takes svṛ "to sound". In thal, iṭ is optional by SK 2279: sasvaritha or sasvartha. For va and ma, SK 2381 follows.
Rule. No iṭ comes to a kit affix after śri. Nor does it come after a one-vowelled stem ending in uk, that is, in u, ṛ or ḷ. "One-vowelled" (ekāc) is carried forward from SK 2246, and "no iṭ" from SK 2981. Bhaṭṭoji's gloss reads "git or kit": an affix marked with g counts as well.
Example. One of Bhaṭṭoji's commentators gives kṛt forms: śritaḥ, bhūtaḥ "become", bhūtvā. The affixes are kit, so no iṭ.
For svṛ in liṭ the rule meets a stronger one:
sasvṛ + va → va is kit, so this rule would bar iṭ
→ but by the krādi restriction (SK 2293), iṭ is obligatory in liṭ for all roots except kṛ and the rest
→ sasvariva "we two sounded". Likewise sasvarima.
Applies when / blocked by. SK 2279 would make iṭ optional for svṛ, and it is the later rule. But this prohibition belongs to a section begun earlier. By the force of that earlier beginning it overrides the option. The krādi restriction then gives iṭ in liṭ all the same. In lṛṭ, SK 2366 wins as the later rule: svariṣyati. Other forms: svaryāt by SK 2380; luṅ asvārīt with iṭ, or asvārṣīt without.
Bhaṭṭoji then lists smṛ "to think of", dvṛ "to cover" and sṛ "to go". sṛ is in the krādi list, so it takes no iṭ in liṭ: sasartha, sasṛva. In āśīrliṅ it takes riṅ (SK 2367): sriyāt. The luṅ is asārṣīt.
The 1915 printed edition spells the first word with a single t, sarti.
Rule. After sṛ "to go", śās and ṛ "to go", aṅ replaces cli in a luṅ that denotes the agent.
Example. The rule reaches only the sṛ and ṛ of the third class, the juhotyādi. So the bhvādi sṛ keeps sic:
a + sṛ + s + ī + t → no aṅ → vṛddhi → asārṣīt "he went".
One of Bhaṭṭoji's commentators gives the true aṅ forms: asarat, aśiṣat, ārat.
Applies when / blocked by. Why only the third class? śās is a root whose śap is deleted (luptaśap). By association (sāhacarya) with śās, sṛ and ṛ are also taken as roots whose śap is deleted. Those are the juhotyādi sṛ and ṛ, not the bhvādi ones.
In the sense of swift going, sṛ takes the substitute dhau before a śit affix, by SK 2360: dhāvati "he runs". Then comes the root ṛ "to go and to reach": ṛcchati, with ṛccha by SK 2360.
Rule. Guṇa comes before liṭ for three things. They are ṛcch of the sixth class (tudādi), the root ṛ, and roots ending in ṝ. "Guṇa" is carried forward from SK 2379, and "liṭ" from SK 2388.
Example. One of Bhaṭṭoji's commentators gives the steps.
ṛ + ṇal → doubling → ṛ ṛ + a
→ the abhyāsa ṛ becomes a, lengthened → ā ṛ + a
→ guṇa with r, then vṛddhi of the penultimate, as under SK 2379 → ā ār + a
→ like vowels merge → āra "he went".
ṛ + atus → ā ṛ + atus → atus is kit, yet this rule gives guṇa → āratuḥ. Likewise āruḥ.
Applies when / blocked by. Without this rule, the kit atus would bar guṇa. The same rule gives teratuḥ from tṝ (SK 2390). The root ṛ is read into the sūtra's plural "ṛcchaty-ṝtām"; one of Bhaṭṭoji's commentators says the plural shows it.
Rule. After ad "to eat", ṛ and vyeñ "to cover", the ending thal always takes iṭ. "thal" is carried forward from SK 2294. The "optionally" of SK 2404 is not carried.
Example.
ṛ + thal → iṭ always → āritha "you went".
The rest of ṛ: artā (luṭ), ariṣyati (lṛṭ), aryāt (āśīrliṅ, guṇa by SK 2380), ārṣīt, ārṣṭām (luṅ).
Applies when / blocked by. Without this rule, SK 2294 (acas tāsvat…) would forbid iṭ in thal for ṛ. One of Bhaṭṭoji's commentators says so.
gṛ "to sprinkle" is not in this rule. So its thal has no iṭ: jagartha. With va it has iṭ by the krādi restriction: jagriva. Its āśīrliṅ takes riṅ: griyāt.
Bhaṭṭoji then gives sru "to go" (one commentator: of liquids, "to flow"). Its luṅ takes caṅ by SK 2312: asusruvat. Here uvaṅ wins over guṇa of the light penultimate. uvaṅ depends on caṅ and is inner (antaraṅga). The guṇa depends on the outer ending t.
Last comes ṣu "to impel": suṣotha or suṣavitha, suṣuviva, sotā. Bhaṭṭoji glosses its sense prasava as "permission".
Rule. After stu, su and dhūñ, sic takes iṭ before a parasmaipada ending.
Example. For the bhvādi ṣu:
a + su + s + ī + t → iṭ to sic → a + su + i + s + ī + t
→ vṛddhi, and sic lost between iṭ and īṭ (SK 2266) → asāvīt "he impelled".
Applies when / blocked by. Some take su here to mean only sunoti, the fifth-class root. They reason from its neighbours stuñ and dhūñ, which both have the marker ñ. On that view the bhvādi su takes no iṭ: asauṣīt. Why "before parasmaipada"? The ātmanepada forms are astoṣṭa and adhoṣṭa.
Next Bhaṭṭoji names śru "to hear".
śru + śap + ti → śṛ + nu + ti → śṛṇotiThree roots take a replacement before certain affixes. śru "to hear" becomes śṛ and takes the affix śnu, giving śṛṇoti (SK 2386, SK 2387). deṅ "to protect" becomes digi in liṭ, giving digye (SK 2388). The ghu roots and sthā take i before taṅ in luṅ, giving adita (SK 2389).
Rule. śru is replaced by śṛ. Also, the affix śnu comes in the place where śap would come. The "also" (ca) brings śnu from SK 2523. So this is an exception (apavāda) to śap.
Example.
śru + śap + ti → śṛ for śru, śnu for śap → śṛ + nu + ti
→ nu takes guṇa before ti → n becomes ṇ after ṛ (a rule not given in this chapter) → śṛṇoti "he hears".
śṛ + nu + tas → no guṇa → śṛṇutaḥ.
Applies when / blocked by. śnu is ṅit by SK 2234, being a sārvadhātuka not marked with p. So the root takes no guṇa.
Watch out. The rule works only where śap would come. That is, before a sārvadhātuka affix in the sense of the agent.
Rule. The u of hu is replaced by yaṇ (a semivowel). So is the u of a many-vowelled stem ending in śnu. This happens before a vowel-initial sārvadhātuka affix. The u must not follow a consonant cluster. The rule is an exception to uvaṅ.
Example.
śṛṇu + anti → anti is ṅit, so no guṇa, and uvaṅ would come
→ this rule gives v for u → śṛṇvanti "they hear".
Other forms of śru: śṛṇomi. śṛṇvaḥ or śṛṇuvaḥ, śṛṇmaḥ or śṛṇumaḥ, with u optionally lost by SK 2333. Also śuśrotha, śuśruva, śṛṇu, śṛṇavāni, śṛṇuyāt, śrūyāt, aśrauṣīt.
Applies when / blocked by. Why "not after a cluster"? āpnuvanti keeps uv. Why "sārvadhātuka"? juhuvatuḥ is liṭ, so no yaṇ. Bhaṭṭoji's commentators give these.
Bhaṭṭoji next lists several roots.
adudruvat.śatrūn jayati "he overcomes his enemies". It also means "to give way": adhyayanāt parājayate "he gives way before study", that is, he wearies of studying. The ātmanepada is by SK 2685.smayate, siṣmiye.gāte serves for three endings, ta, ātām and jha.ūve. uvaṅ comes before the merging of like vowels. The reason is a paribhāṣā, a rule of interpretation: "an operation on the stem is stronger than one on the sounds".praṇimayate, with ṇ by SK 2285.dayate.Rule. Before liṭ, deṅ "to protect" (dayati) is replaced by digi. The Vṛtti, an older commentary, says this substitute is meant to block the doubling.
Example.
de + e (liṭ, ātmanepada) → digi for the root → digy + e (i becomes y before the vowel e)
→ no doubling → digye "he protected".
Applies when / blocked by. One of Bhaṭṭoji's commentators explains the blocking. Before liṭ both doubling and digi are due. digi is taught specially for this case, so it sets the doubling aside.
Watch out. Do not write didigye. digi replaces the root, and no doubling follows.
Rule. For sthā and the ghu roots, i replaces the final. Also, sic is kit. Both happen before a taṅ ending. "sic" is carried from SK 2697, and "kit" from SK 2242.
Example.
a + dā + s + ta → i for ā, and sic is kit → a + di + s + ta
→ sic lost before the jhal t, the stem ending in short i (SK 2369) → adita "he gave (for himself)".
Likewise adithāḥ and adiṣi.
Applies when / blocked by. The i takes no guṇa: the ending is ṅit, and sic is kit. Before ātām no jhal follows, so sic stays: adiṣātām. For sthā: asthita. Bhaṭṭoji's commentators note that the rule cannot act before parasmaipada. There SK 2223 has already removed sic.
Bhaṭṭoji ends the ṅit roots here. pyaiṅ "to grow" gives papye. pūṅ "to purify" gives pavate, pupuve. ḍīṅ "to fly" gives ḍayate, ḍiḍye, ḍayitā. Then comes tṝ "to float, to cross".
tṝ → tarati, teratuḥ, tarītā / taritā, tīryāt, atāriṣṭāmThe root tṝ "to cross" ends in long ṝ. Such roots have three habits. Their ṝ becomes i (SK 2390), though guṇa and vṛddhi take precedence: tarati. Their iṭ may be long: tarītā (SK 2391). But not before the sic of the parasmaipada luṅ: atāriṣṭām (SK 2392).
Rule. A stem that is a root ending in ṝ takes i in place of the ṝ. The i is followed by r, as with ṛ (raparatva).
A vārttika follows, a supplementary statement on the rule: "guṇa and vṛddhi, by vipratiṣedha, set aside i and u". Vipratiṣedha is the principle that of two conflicting rules the later one wins (SK 175).
Example.
tṝ + śap + ti → i by this rule, or guṇa by the rule for sārvadhātuka affixes?
→ by the vārttika, guṇa wins → ar → tarati "he crosses".
In liṭ: tṝ + atus → doubling → tatṝ + atus → guṇa (SK 2383) → tatar + atus → e, abhyāsa lost (SK 2301) → teratuḥ. Likewise teruḥ.
Applies when / blocked by. The i is an inner (antaraṅga) operation. So it might claim to go first. Bhaṭṭoji's commentators say the vārttika was stated to prevent this. Why "a root"? mātṝṇām "of mothers" keeps its ṝ.
Watch out. The vārttika also covers the u of another rule, SK 2494. It is not about tṝ alone.
Rule. After vṛ (both vṛṅ and vṛñ) and after a root ending in ṝ, iṭ is optionally long. But not in liṭ. "iṭ" is carried forward from SK 2184, and "long, not in liṭ" from SK 2562.
Example.
tṝ + iṭ + tā → guṇa of the root, iṭ long or short
→ tarītā or taritā "he will cross".
Why "not in liṭ"? The liṭ has only teritha.
In āśīrliṅ the affix is kit, so there is no guṇa:
tṝ + yāt → ir by SK 2390 → lengthening by SK 354 (hali ca) → tīryāt.
Rule. Before a sic followed by parasmaipada endings, the iṭ after these roots is not lengthened. The "not" is carried forward from SK 2529.
Example.
a + tṝ + iṭ + s + tām → SK 2391 would lengthen the iṭ; this rule forbids it
→ atāriṣṭām "they two crossed".
Bhaṭṭoji then announces eight roots marked with the anudātta accent on their marker vowel (anudāttet). Such roots take the ātmanepada. The first four are gup "to protect", tij "to sharpen", mān "to honour" and badh "to bind".
gup + san → jugupsate ("censures")The affix san usually makes a desiderative, "wishes to do". Here it is different. After seven roots it comes in its own sense, and the sense is fixed by a vārttika. So gup with san means "to censure": jugupsate. SK 2393 and SK 2394 add the san. SK 2395 doubles the root.
Rule. The affix san comes after gup, tij and kit.
Bhaṭṭoji gives this sūtra no gloss of its own. He explains it together with the next, in one sentence: "after the roots stated in the two sūtras, san comes". gup and tij are in the list just read (SK 2392). kit comes a little later, among the parasmaipada roots (SK 2407).
Example.
gup + san → gupsa → doubling by SK 2395 → jugupsa → jugupsate "he censures".
Applies when / blocked by. The senses in which this san comes are given as vārttikas under SK 2394.
Rule. san also comes after mān, badh, dān and śān. And for these four the i of the abhyāsa is made long. That i is the one given to the abhyāsa before san by SK 2317 (sanyataḥ).
Seven vārttikas then fix the senses:
By SK 2304, sanādyantā dhātavaḥ, the form ending in san is itself a root.
Example.
mān + san → doubling (SK 2395) → i in the abhyāsa (SK 2317) → mi mān sa
→ this rule lengthens the i → mī mān sa
→ n becomes anusvāra before s → mīmāṃsate "he inquires".
Applies when / blocked by. Bhaṭṭoji's commentators say these senses come from the Vṛtti and other older works. The root lists give gup as "to protect". One of Bhaṭṭoji's commentators says such sense-labels only illustrate; a root may have other senses.
Rule. In a form ending in san or in yaṅ, the first single-vowelled part is doubled. If the root begins with a vowel, the second part is doubled. This rests on the headings SK 2175 and SK 2176. Then the usual operations on the abhyāsa follow.
Example.
gup + sa → doubling → gup gup sa → the abhyāsa keeps its first consonant → gu gup sa
→ g of the abhyāsa becomes j → jugupsa → jugupsate "he censures".
Likewise jugupsāṃcakre (liṭ with ām), titikṣate, mīmāṃsate.
For badh, the root is doubled as badh, before its sound changes count:
badh + sa → ba badh sa → i, lengthened by SK 2394 → bī badh sa
→ b becomes bh, dh becomes t → bībhatsate (in the sense of a change of mind).
Applies when / blocked by. The case in sanyaṅoḥ is a part-whole genitive: "of what ends in san". It does not mean "before san". Bhaṭṭoji's commentators say so. In pratīṣiṣati and aṭāṭyate the doubled part takes in sounds of the affix. That works only if the rule speaks of the whole form.
Bhaṭṭoji adds four points about these roots.
The anudāttet list goes on with rabh "to be hasty"; with ā it means "to begin": ārabhate, ārebhe, rabdhā. Then labh "to obtain": labhate. Then svañj "to embrace".
Watch out. Do not add iṭ to this san: jugupsate, not jugupiṣate. The reason is the last point above.
svañj + śap + te → svajate; pari- → pariṣvajate; liṭ pariṣasvajeSome roots lose their nasal before śap: svañj "to embrace" gives svajate (SK 2396, SK 2397). The same roots raise questions about ṣ after a prefix. svañj has pariṣvajate, but in liṭ pariṣasvaje. skand "to go" has an optional ṣ after vi and pari (SK 2398, SK 2399).
Rule. The n of daṃś "to bite", sañj "to cling" and svañj is deleted before śap. "Deletion of n" is carried forward from SK 2544.
Bhaṭṭoji gives this sūtra no gloss of its own. He glosses it together with the next, SK 2397.
Example.
svañj + śap + te → n (here ñ) deleted before śap
→ svaj + a + te → svajate "he embraces".
Bhaṭṭoji later cites this rule for daśati "he bites" and sajati "he clings".
Watch out. The loss is before śap only. In svaṅktā (luṭ) the nasal stays.
Rule. The n of rañj "to colour" is also deleted before śap. Bhaṭṭoji's joint gloss of the two sūtras: "of these, loss of n before śap".
Example.
svañj + śap + te → svajate; with pari, pariṣvajate.
rañj + śap + ti → rajati "he colours" (given under the root rañj, SK 2407).
Applies when / blocked by. Bhaṭṭoji next cites a rule of another grammar. "For śranth, granth, dambh and svañj, liṭ is optionally kit." The Bhāṣya gives two examples of it, debhatuḥ and sasvaje. On the strength of these, the whole rule is accepted here. When liṭ is kit, the n is lost: sasvaje, beside sasvañje.
Then a vārttika on SK 2361: svañjer upasaṃkhyānam, "svañj is to be added". So, as with sad, the s after the abhyāsa takes no ṣ. The s of the abhyāsa still does:
pari + sasvaje → pariṣasvaje or pariṣasvañje.
Other forms: sasvajiṣe, svaṅktā, svaṅkṣyate, svajeta, svaṅkṣīṣṭa.
After prati, the ṣ comes across the augment a by SK 2276: pratyaṣvaṅkta. After pari, ni and vi it is optional by SK 2359: paryaṣvaṅkta or paryasvaṅkta. Bhaṭṭoji says this is why svañj is named again in SK 2275.
Bhaṭṭoji then lists hada "to void the bowels": hadate, jahade, hattā, ahatta. Then the parasmaipada roots begin, with skandir "to go, to dry". Its forms include caskanditha or caskantha, skantā, and skadyāt with loss of n. Being marked with ir, it takes aṅ optionally in luṅ: askadat.
Rule. After vi, the s of skand is optionally replaced by ṣ. This does not happen before a niṣṭhā, the past participle affix kta or ktavatu. "Optionally" is carried forward from SK 2359, under the heading SK 210.
Example.
vi + skand + tṛc (tṛc, the affix of agent nouns) → optional ṣ → viṣkantā or viskantā, agent nouns.
Before a niṣṭhā: viskannaḥ only.
Applies when / blocked by. Bhaṭṭoji says the rule applies only before a kṛt, an affix that makes a noun from a root. The reason is the wording "not before a niṣṭhā". This kind of negation (paryudāsa) points to something of the same kind as what it excludes. So it means "before a kṛt other than a niṣṭhā". A finite verb like viskandati keeps its s; scholars discuss this.
Watch out. skand is not taught with ṣ. So this ṣ would not come at all without the rule.
Rule. After pari too, the s of skand is optionally replaced by ṣ. The rule is split off from SK 2398 (yogavibhāga). So its "not before a niṣṭhā" does not apply here.
Example.
pari + skandati → pariṣkandati or pariskandati.
Before a niṣṭhā: pariṣkaṇṇaḥ or pariskannaḥ.
Applies when / blocked by. In the ṣ option, the n becomes ṇ: pariṣkaṇṇaḥ. Could the ṣ be treated as not yet there for this purpose? It depends on two words, the prefix and the root. Bhaṭṭoji says no. An operation between root and prefix counts as inner. The Bhāṣya says: "first the root is joined with the prefix, then with the means (sādhana)". Here the means is the affix. On the other view, the order is reversed, and there is no ṇ.
Bhaṭṭoji then lists yabh: yebhitha or yayabdha, yabdhā, ayāpsīt. Then ṇam "to bow": nemitha or nanantha, nantā, anaṃsīt (with sak by SK 2377). Then gamḷ and sṛpḷ, both "to go".
dṛś → paśyati, dadarśitha / dadraṣṭha, draṣṭā, adarśat / adrākṣītThis group follows a run of common roots: gam and sṛp, both "to go", yam "to cease", tap "to heat" and dṛś "to see". Its model is dṛś. Before śap it takes the stem paśya: paśyati (SK 2360). In thal the iṭ is optional: dadarśitha or dadraṣṭha (SK 2404). Without iṭ it takes the augment am: dadraṣṭha, draṣṭā (SK 2405). In luṅ it has two forms. With aṅ it takes guṇa: adarśat (SK 2406). Without aṅ it keeps sic: adrākṣīt (SK 2407).
Rule. iṣ (the sixth-class root, cited as iṣu), gam and yam take ch for their final. This happens before a śit affix. "Before a śit" is carried forward from SK 2320.
Example.
gam + śap + ti → śap is śit → ch for m → gach + a + ti
→ a c is added before ch (that rule is not given here) → gacch + a + ti → gacchati "he goes".
Outside the śit domain the root keeps m: jagāma. jagmatuḥ and jagmuḥ lose the penultimate a by SK 2363. jagamitha or jagantha; gantā.
yam gives yacchati by the same rule (see SK 2402).
Applies when / blocked by. One of Bhaṭṭoji's commentators reports a further limit from the Bhāṣya. The śit affix must begin with a vowel. So iṣyati has no ch.
Rule. After gam, an ārdhadhātuka affix beginning with s takes iṭ. This happens before parasmaipada endings. "Beginning with s" is carried forward from SK 2506.
Example.
gam + sya + ti → sya begins with s, and the ending is parasmaipada
→ iṭ → gam + i + ṣya + ti → gamiṣyati "he will go".
Applies when / blocked by. In luṅ, gamḷ is marked with ḷ, so cli becomes aṅ (SK 2343). SK 2363 would delete its penultimate a before a kit or ṅit affix. But that rule says "not before aṅ". So the a stays: agamat "he went".
One of Bhaṭṭoji's commentators reads "parasmaipada" as "where there is no ātmanepada". So even the agent noun saṃjigamiṣitā gets the iṭ.
Next comes sṛp "to go": sarpati, sasarpa.
Rule. A root that is anudātta in its taught form and has ṛ as its penultimate optionally takes the augment am. This happens before a jhal-initial affix that is not kit. Such a root takes no iṭ (SK 2246). The augment am is marked with m, so it goes after the last vowel. "am before a jhal-initial non-kit affix" is carried forward from SK 2405.
Example.
sṛp + tā → am after the ṛ → sṛ a p + tā → ṛ becomes r → sraptā.
Without am: guṇa of the penultimate → sarptā. Both mean "he will go".
Likewise srapsyati or sarpsyati. The luṅ has aṅ: asṛpat.
Applies when / blocked by. One of Bhaṭṭoji's commentators gives counter-examples. kṛ has no penultimate ṛ: kartā. The ending of sasarpa does not begin with a jhal. The affix of sṛptaḥ is kit.
Bhaṭṭoji then gives yam "to cease": yacchati (SK 2400), yemitha or yayantha, yantā. Its luṅ takes sak and iṭ by SK 2377: ayaṃsīt, ayaṃsiṣṭām. Then tap "to heat": taptā, atāpsīt.
Rule. The s of the prefix nis is replaced by ṣ before tap. This does not happen in the sense of repetition (āsevana).
Example.
nis + tapati → ṣ for s → niṣ + tapati → t becomes ṭ after ṣ → niṣṭapati "he heats thoroughly".
With repetition: nistapati "he heats again and again".
Applies when / blocked by. One of Bhaṭṭoji's commentators says the ṣ would not come otherwise. The s of nis is not a substitute, and it stands at the end of a word.
Bhaṭṭoji then gives tyaj "to abandon": tatyajitha or tatyaktha, tyaktā, atyākṣīt. Then ṣañj "to cling", which loses its n before śap by SK 2396: sajati, saṅktā. Then dṛś "to see": paśyati, with paśya by SK 2360.
Rule. After sṛj and dṛś, the ending thal optionally takes iṭ. "thal" is carried forward from SK 2294, and "iṭ" from SK 2401.
Example.
dadṛś + thal → with iṭ → dadarśitha "you saw".
Without iṭ, SK 2405 adds am → dadraṣṭha.
Applies when / blocked by. One of Bhaṭṭoji's commentators says the option matters because the krādi restriction (SK 2293) would otherwise make iṭ obligatory.
Rule. sṛj and dṛś take the augment am before an affix that begins with a jhal and is not kit.
Example.
dadṛś + tha → tha begins with a jhal, and thal is not kit
→ am after the ṛ → dadṛ a ś + tha → ṛ becomes r → ś becomes ṣ, th becomes ṭh
→ dadraṣṭha "you saw".
With iṭ the affix begins with i, so no am: dadarśitha.
Likewise draṣṭā "he will see" and drakṣyati.
In āśīrliṅ the affix begins with y, so no am: dṛśyāt.
Applies when / blocked by. For sṛj and dṛś, SK 2402 would give am only optionally. One of Bhaṭṭoji's commentators says this rule makes it obligatory. Why "a jhal"? sasarja, dadarśa. Why "not kit"? sṛṣṭaḥ, dṛṣṭaḥ.
dṛśir is marked with ir, so in luṅ aṅ is optional. The next two rules give both options.
Rule. Before aṅ, guṇa comes for a stem ending in ṛ or ṝ, and for dṛś.
Example.
a + dṛś + aṅ + t → aṅ is ṅit, so guṇa would be barred
→ this rule gives guṇa → adarśat "he saw".
Applies when / blocked by. The word "before aṅ" in this rule is carried forward into SK 2421, śvayater aḥ. For the option without aṅ, see SK 2407.
Rule. After dṛś, cli is not replaced by ksa. "cli" is carried forward from SK 2222, and "ksa" from SK 2336. SK 2336 would otherwise give ksa: dṛś ends in ś, has ṛ before it, and takes no iṭ.
Example. The option without aṅ:
a + dṛś + cli + t → no ksa, so sic stays
→ am by SK 2405, ṛ becomes r, vṛddhi → ś becomes k, s becomes ṣ
→ adrākṣīt "he saw".
Applies when / blocked by. A vārttika follows here. After spṛś, mṛś, kṛṣ, tṛp and dṛp, cli becomes sic optionally. kṛṣ means "to draw (a furrow)". So kṛṣ has three luṅ forms:
akrākṣīt, akrāṣṭām;akārkṣīt, akārṣṭām, akārkṣuḥ;akṛkṣat, akṛkṣatām, akṛkṣan.Bhaṭṭoji then works through many roots. The main points:
daśati (SK 2396), dadaṃśitha or dadaṃṣṭha, daṃṣṭā, daśyāt, adāṅkṣīt.kraṣṭā or karṣṭā, with optional am by SK 2402.dehitha or dadagdha, dagdhā, dhakṣyati, adhākṣīt.mimeha, meḍhā, amikṣat.cikitsati "he treats (a disease)", with san by SK 2393. In the sense of dwelling it is a tenth-class root: ketayati.dīdāṃsati, dīdāṃsate, śīśāṃsati, śīśāṃsate. Elsewhere dānayati.pacati, pacate, pecitha or papaktha, paktā.bhejatuḥ, bhejitha or babhaktha, abhākṣīt.rajati, rajate (SK 2397), arāṅkṣīt.śaśāpa, śepe, aśāpsīt.titviṣe, atvikṣat.yajati, yajate. It opens the next group.yaj + ṇal → yayāj + a → iyāja; yaj + atus → ij + atus → ījatuḥThe last roots of the bhvādi are the yajādi, "yaj and the rest". Bhaṭṭoji's commentators list them in a verse: yaj, vap, vah, vas, veñ, vyeñ, hveñ, vad and śvi. In liṭ and before kit affixes, their semivowel turns into its vowel, by saṃprasāraṇa: y to i, v to u. Two rules divide the work. Before a kit affix the whole root changes, then doubles: ījatuḥ (SK 2409). Before a non-kit liṭ ending, only the abhyāsa changes: iyāja (SK 2408). The rest of the group handles vas, veñ, vyeñ and hveñ.
Rule. In liṭ, the abhyāsa of "both" groups takes saṃprasāraṇa. The two groups are those of SK 2409 (vac, svap and the yajādi) and of SK 2412 (grah and the rest). "Saṃprasāraṇa" is carried forward from SK 1003.
Example.
yaj + ṇal → doubling, the abhyāsa keeping its first consonant → ya yāj + a (vṛddhi before ṇal)
→ y of the abhyāsa becomes i, and i + a gives i (SK 330) → iyāja "he worshipped".
Applies when / blocked by. ṇal is not kit, so SK 2409 does not act. Only this rule changes the abhyāsa. The same holds in thal: iyajitha or iyaṣṭha.
Watch out. Why "in liṭ"? In the san-forms vivakṣati and yiyakṣati the abhyāsa does not change. One of Bhaṭṭoji's commentators gives these.
Rule. vac "to speak", svap and the yajādi roots take saṃprasāraṇa before a kit affix.
Example.
yaj + atus → this rule is later than doubling, so it acts first → ij + atus
→ doubling still comes, since a blocked rule may apply again (punaḥprasaṅga) → ij ij + atus
→ the abhyāsa keeps only an initial consonant; ij begins with a vowel, so its j goes → i ij + atus → like vowels merge → ījatuḥ "they two worshipped".
Likewise ījuḥ, and the ātmanepada īje.
In āśīrliṅ the augment yāsuṭ is kit: ijyāt.
Other forms of yaj: yaṣṭā, yakṣyati, yakṣīṣṭa, ayākṣīt, ayaṣṭa.
Applies when / blocked by. A kit liṭ ending falls under this rule. A non-kit one falls under SK 2408. One of Bhaṭṭoji's commentators states the division.
Bhaṭṭoji then gives vap "to sow". It also means "to cut": keśān vapati "he cuts (shaves) the hair". Forms: uvāpa, ūpe, upyāt, praṇyavāpsīt.
Then vah "to carry": uvāha, ūhe, voḍhā, avākṣīt. In thal with iṭ: uvahitha. Without iṭ, SK 2357 gives o for a after the loss of ḍh: uvoḍha.
Last vas "to dwell", a parasmaipada root: vasati, uvāsa.
Rule. The s of śās, vas and ghas is replaced by ṣ. This happens after a sound of the iṇ class or a k-class consonant. "The s" is carried forward from SK 335. SK 211 and SK 210 are the headings.
Example.
vas + atus → saṃprasāraṇa (SK 2409) and doubling → ū s + atus
→ the s follows ū, so it becomes ṣ → ūṣatuḥ "they two dwelt". Likewise ūṣuḥ.
Other forms: uvasitha or uvastha, vastā.
Before an ārdhadhātuka affix in s, SK 2342 turns the root's s into t: vatsyati, avātsīt, avāttām.
In āśīrliṅ: uṣyāt.
Applies when / blocked by. Without this rule there would be no ṣ. The s of vas belongs neither to a substitute nor to an affix. One of Bhaṭṭoji's commentators says so.
Bhaṭṭoji then gives veñ "to weave": vayati, vayate.
Rule. In liṭ, veñ is optionally replaced by vay. The i of vayi is only for pronunciation. "Optionally" is carried forward from A 2.4.40 (SK 2424).
Example.
veñ + ṇal → vay → doubling → the abhyāsa takes saṃprasāraṇa (SK 2408), v → u → u vay + a
→ vṛddhi of the penultimate → uvāya "he wove".
Applies when / blocked by. Had the i been a marker, the root would take num. So it is not a marker. One of Bhaṭṭoji's commentators adds that uvāya could be made without vay. The substitute is needed for forms like ūyatuḥ (SK 2413).
Rule. Nine roots take saṃprasāraṇa before a ṅit affix, and also before a kit one. They are grah, jyā, vay, vyadh, vaś, vyac, vraśc, pracch and bhrasj. The "also" brings in "kit" from SK 2409.
Example. One of Bhaṭṭoji's commentators gives gṛhṇāti, gṛhītaḥ from grah, and jināti, jīnaḥ from jyā.
Applies when / blocked by. vay is named here only for clarity. As the substitute of veñ, it is already a yajādi root. Bhaṭṭoji's commentators say so.
Now take vay + atus. vay has two semivowels, v and y. The last one takes saṃprasāraṇa first; SK 363 shows this. So the y would become i. The next rule stops that.
Rule. In liṭ, the y of vay does not take saṃprasāraṇa. "Not saṃprasāraṇa" is carried forward from SK 363.
Example.
vay + atus → the y is protected → the v takes saṃprasāraṇa → uy + atus
→ doubling, like vowels merge → ūyatuḥ "they two wove". Likewise ūyuḥ.
Rule. In a kit liṭ, that y of vay is optionally replaced by v. "y of vay in liṭ" is carried forward from SK 2413.
Example.
ūyatuḥ → v for y → ūvatuḥ. So ūyatuḥ or ūvatuḥ; ūyuḥ or ūvuḥ.
In thal: uvayitha. thal is not kit, so the y stays.
The ātmanepada: ūye or ūve.
Applies when / blocked by. Why is iṭ obligatory in uvayitha? vay is taught for liṭ only, so it never meets tās. The rules that forbid iṭ in thal look at a root's behaviour before tās. They cannot reach vay. So the krādi restriction gives iṭ.
Why the ātmanepada? vay stands in place of veñ, which is marked with ñ. By sthānivadbhāva a substitute is treated like what it replaces. So vay counts as ñit, and takes both sets of endings.
Rule. Where vay does not replace it, veñ takes no saṃprasāraṇa in liṭ. "Not saṃprasāraṇa" is carried forward from SK 363, and "in liṭ" from SK 2413.
Example.
ve + ṇal → no saṃprasāraṇa → ā for e (SK 2370) → doubling → vavā + a
→ au for ṇal (SK 2371) → vavau "he wove".
vavā + atus → ā lost (SK 2372) → vavatuḥ. Likewise vavuḥ, vavitha or vavātha, vave.
Outside liṭ saṃprasāraṇa returns: ūyāt (āśīrliṅ, kit). Also vātā, vāsīṣṭa, avāsīt.
Applies when / blocked by. "Kit" is not carried forward here. So the rule covers ṇal too, and gives vavau.
Bhaṭṭoji then gives vyeñ "to cover": vyayati.
Rule. In liṭ, vyeñ does not take the ā of SK 2370.
Example.
vye + ṇal → no ā → vṛddhi of the final e → vyai
→ doubling → the y of the abhyāsa becomes i (SK 2408) → vi vyai + a
→ vivyāya "he covered". Likewise vivyatuḥ, vivyuḥ.
In thal, iṭ is obligatory by SK 2384: vivyayitha.
The uttama singular: vivyāya or vivyaya. The ātmanepada: vivye.
Outside liṭ the ā returns: vyātā, vyāsīṣṭa, avyāsīt, avyāsta. In āśīrliṅ: vīyāt.
Applies when / blocked by. The rule halādiḥ śeṣaḥ keeps only the first consonant of an abhyāsa, here v. Then the v would take saṃprasāraṇa, giving the wrong uvyāya. Bhaṭṭoji explains why that does not happen. The word "both" in SK 2408 is not needed to include the two groups, since they would come in anyway. So the word has a force (sāmarthya) of its own. By that force the saṃprasāraṇa sets aside halādiḥ śeṣaḥ, though that is the later rule. One of his commentators explains this as a restriction (niyama): in the abhyāsa of these roots, saṃprasāraṇa happens and nothing else.
Next comes hveñ "to call", in rivalry and in sound.
Rule. hveñ takes saṃprasāraṇa when it is about to become abhyasta, that is, about to be doubled. After that it is doubled. "hveñ takes saṃprasāraṇa" is carried forward from SK 2586. One of Bhaṭṭoji's commentators treats this sūtra as split off from that one.
Example.
hve + ṇal → ā (SK 2370) → saṃprasāraṇa → hu → doubling → hu hu + a
→ h of the abhyāsa becomes j → ju hu + a → vṛddhi, āv → juhāva "he called".
Likewise juhuvatuḥ, juhuvuḥ, juhotha or juhavitha, juhuve.
Outside liṭ: hvātā, hūyāt, hvāsīṣṭa.
Applies when / blocked by. Why "about to become" and not "having become"? If saṃprasāraṇa came only after the doubling, SK 363 would block the saṃprasāraṇa of the abhyāsa. One of Bhaṭṭoji's commentators gives this reason.
hvā → ahvat / ahvata / ahvāsta; śvi → aśvat / aśiśviyat / aśvayītTwo yajādi roots have unusual luṅ forms. hvā (from hveñ) takes aṅ, always in the parasmaipada and optionally in the ātmanepada (SK 2418, SK 2419). śvi "to go, to grow" has three luṅ forms: with aṅ, with caṅ, and with sic (SK 2421). It also has an optional saṃprasāraṇa in liṭ (SK 2420). With śvi the bhvādi roots end.
Rule. After lip "to smear", the root sic "to sprinkle" (not the affix sic) and hvā, aṅ replaces cli. "aṅ" is carried forward from SK 2438, and "cli" from SK 2222.
Example. One of Bhaṭṭoji's commentators gives alipat "he smeared" and asicat "he sprinkled".
For hvā, see SK 2419.
Applies when / blocked by. One of Bhaṭṭoji's commentators says this rule is obligatory in the parasmaipada. The next rule gives an option for the ātmanepada.
Rule. Before ātmanepada endings, the aṅ of SK 2418 is optional.
Example.
a + hvā + aṅ + t → aṅ is vowel-initial and ṅit, so the ā of hvā is lost by SK 2372
→ ahvat "he called". Likewise ahvatām, ahvan.
Ātmanepada with aṅ: ahvata. Without aṅ: ahvāsta.
Applies when / blocked by. The parasmaipada forms have aṅ always, by SK 2418. Only the ātmanepada has the choice.
Bhaṭṭoji then gives vad "to speak clearly". accha vadati means "he speaks facing (someone)". Liṭ: uvāda (SK 2408), ūdatuḥ (SK 2409), uvaditha. Also vaditā, udyāt. In luṅ, vṛddhi comes by SK 2267: avādīt. Last comes śvi "to go, to grow": śvayati.
Rule. śvi optionally takes saṃprasāraṇa in liṭ and before yaṅ. "Saṃprasāraṇa" is carried forward from SK 1003, and "in liṭ and before yaṅ" from SK 2327.
Example. With saṃprasāraṇa:
śvi + ṇal → v becomes u, which absorbs the i (SK 330) → śu → doubling → śu śu + a → vṛddhi, āv → śuśāva. Likewise śuśuvatuḥ.
A vārttika follows: for śvi in liṭ, the saṃprasāraṇa due to the abhyāsa is forbidden. So SK 2408 does not act. Without saṃprasāraṇa:
śvi + ṇal → doubling → śi śvi + a → vṛddhi, āy → śiśvāya. Likewise śiśviyatuḥ.
Other forms: śvayitā, śvayet, śūyāt. In luṅ aṅ is optional by SK 2291.
Applies when / blocked by. Without the vārttika, SK 2408 would change the abhyāsa and give śuśvāya. Bhaṭṭoji's commentators both say so. Before yaṅ, they give śośūyate or śeśvīyate.
Rule. Before aṅ, the final i of śvi is replaced by a. "Before aṅ" is carried forward from SK 2406.
Example. With aṅ:
a + śvi + a + t → a for i → a + śva + a + t
→ the two a's give one a (pararūpa) → aśvat "he grew". Likewise aśvatām, aśvan.
With caṅ by SK 2375, doubling and iyaṅ: aśiśviyat.
With neither, sic comes, and SK 2299 forbids vṛddhi: aśvayīt.
Applies when / blocked by. Why pararūpa and not a long ā? The rule for like vowels would merge the two into ā. Bhaṭṭoji names pararūpa to prevent that. One of Bhaṭṭoji's commentators adds that a-deletion does not apply either: the stem was not taught ending in a.
Then Bhaṭṭoji writes vṛt, "finished": the yajādi roots are finished. He does not say that the bhvādi is finished. The reason is that the bhvādi is an ākṛtigaṇa, a class defined by its type, not a closed list. So roots like culump "to delete", in culumpati, belong to it too. Bhaṭṭoji closes: "here end the bhvādi roots of the tiṅanta section".
ṛt + īyaṅ → ṛtīya → ṛtīyateOne sūtra remains. It teaches an affix, īyaṅ, for a root found only in the sūtras. The new stem ṛtīya is itself a root and is conjugated like the bhvādi roots: ṛtīyate.
Rule. After the root ṛt, the affix īyaṅ comes, in the root's own sense. ṛt is a sautra root: it is not in the root list, only in the sūtras. Many say it means disgust (jugupsā); some add pity (kṛpā).
Example.
ṛt + īyaṅ → ṛtīya → a root by SK 2304 (sanādyantā dhātavaḥ)
→ ṛtīya + śap + te → ṛtīyate "he feels disgust".
In liṭ, with ām: ṛtīyāṃcakre.
Applies when / blocked by. When an ārdhadhātuka affix is intended, īyaṅ is optional by SK 2305. Without it, the bare root takes the parasmaipada by SK 2159: ānarta (liṭ), artiṣyati (lṛṭ), ārtīt (luṅ).
Watch out. The root is ṛt, ending in t, not ṛti. Bhaṭṭoji cites it as ṛti, with the i used in naming roots. Bhaṭṭoji's commentators agree on this. They differ on one argument for it; scholars discuss this.
With this rule the bhvādi chapter closes. Its roots ended with śvi and the word vṛt in SK 2421, and SK 2422 adds the one derived root ṛtīya. In the printed edition of the Kaumudī, ṛter īyaṅ already stands at the head of the next section. Either way, the second class of roots, the adādi, "ad and the rest", comes next. There śap is deleted after the root. It is the subject of chapter 45.
abhūt (SK 2225).edhiṣīdhvam; with luṅ and liṭ, one of the three whose dh may become ḍh (SK 2247).ācāmati (SK 2320).roṣṭā) or dh → ḍh after ḍh (SK 2335, SK 2356, SK 2365).krathaṃkratham; taught with the kṛt affixes.ghaṭā, vyathā).mriyate (SK 2367).| SK | what it does | example |
|---|---|---|
| 2151 | laṭ after a root whose action is present | bhū + laṭ → bhū + l |
| 2152 | lakāras denote agent or object (sakarmaka), agent or bare action (akarmaka) | bhavati, agent sense |
| 2153 | heading: "in place of l" | l of bhū + l is replaced |
| 2154 | the eighteen tiṅ replace l | bhū + l → bhū + tip |
| 2155 | substitutes of l are parasmaipada | tip … mas |
| 2156 | taṅ and āna are ātmanepada | ta … mahiṅ, śānac, kānac |
| 2157 | ātmanepada after anudātta-it and ṅit roots | not bhū |
| 2158 | ātmanepada after svarita-it and ñit roots, fruit to the agent | not bhū |
| 2159 | parasmaipada after the rest, in the agent sense | bhū + (tip … mas) |
| 2160 | the triads are prathama, madhyama, uttama | tip tas jhi = prathama |
| 2161 | in each triad: singular, dual, plural | tip = prathama singular |
| 2162 | madhyama with co-referent yuṣmad | tvaṃ bhavasi |
| 2163 | in mockery, madhyama and uttama swap; man singular | ehi manye odanaṃ bhokṣyase |
| 2164 | uttama with co-referent asmad | ahaṃ bhavāmi |
| 2165 | prathama elsewhere | bhū + ti |
| 2166 | tiṅ and śit affixes after a root are sārvadhātuka | ti, śap |
| 2167 | śap before a sārvadhātuka of the agent | bhū + a + ti |
| 2168 | guṇa of a final ik before sārvadhātuka/ārdhadhātuka | bhavati, bhavataḥ |
| 2169 | jh of an affix → ant | bhavanti |
| 2170 | a of the aṅga lengthened before a yañ-initial sārvadhātuka | bhavāmi, bhavāvaḥ, bhavāmaḥ |
| 2171 | liṭ for the past not of today, beyond perception | bhū + liṭ |
| 2172 | endings of liṭ are ārdhadhātuka only; no śap | bare bhū |
| 2173 | parasmaipada endings of liṭ → ṇal, atus, us … | bhū + a |
| 2174 | bhū takes vuk before a vowel of luṅ, liṭ | bhūv + a |
| 2175 | heading: two utterances of the first ekāc | — |
| 2176 | heading: in a vowel-initial root, the second ekāc | — |
| 2177 | before liṭ, the root's first ekāc is doubled | bhūv bhūv a |
| 2178 | the former copy is abhyāsa | [bhūv] bhūv a |
| 2179 | in the abhyāsa only the initial consonant remains | bhū bhūv a |
| 2180 | the abhyāsa vowel is shortened | bhu bhūv a |
| 2181 | a for the u of bhū's abhyāsa before liṭ | bha bhūv a |
| 2182 | in the abhyāsa, jhaś → jaś, khay → car | babhūva |
| 2183 | heading: ābhīya operations asiddha for each other; vārttika makes vuk siddha for uvaṅ | babhūva, not babhova |
| 2184 | iṭ before a val-initial ārdhadhātuka | babhūvitha, babhūviva, babhūvima |
| 2185 | luṭ for the future not of today | bhū + luṭ |
| 2186 | sya before lṛ, tās before luṭ | bhū + tās + ti |
| 2187 | other affixes after a root are ārdhadhātuka | tās takes iṭ |
| 2188 | prathama of luṭ → ḍā, rau, ras; ṭi drops before ḍā | bhavit ā |
| 2189 | guṇa of a light penultimate ik | due on the i of bhavit |
| 2190 | no guṇa or vṛddhi for dīdhī, vevī, iṭ | bhavitā |
| 2191 | s of tās and as lost before an s-initial affix | bhavitāsi |
| 2192 | the same for tās before an r-initial affix | bhavitārau, bhavitāraḥ |
| 2193 | lṛṭ for the future | bhaviṣyati |
| 2194 | loṭ in the senses vidhi and the rest | bhū + loṭ |
| 2195 | liṅ and loṭ also in benediction | bhavatu, bhavatāt |
| 2196 | i of a loṭ ending → u | bhavatu, bhavantu |
| 2197 | tu, hi → tāt in benediction, optionally | bhavatāt |
| 2198 | loṭ takes the substitutes taught for laṅ | bhavatām |
| 2199 | ṅit lakāra: tas, thas, tha, mip → tām, tam, ta, am | bhavatām, bhavatam, bhavata |
| 2200 | s of a ṅit uttama ending always lost | bhavāva, bhavāma |
| 2201 | si of loṭ → hi, not pit | bhava + hi, stuhi |
| 2202 | hi after a is deleted | bhava |
| 2203 | mi of loṭ → ni | bhava + ni |
| 2204 | uttama of loṭ takes pit āṭ | bhavāni, bhavāva, bhavāma |
| 2205 | Laṅ for a past not of today | bhū → abhavat |
| 2206 | The aṅga takes the udātta augment aṭ before luṅ, laṅ, lṛṅ | bhava + ti → a + bhava + ti |
| 2207 | An i-final parasmaipada ending of a ṅit lakāra loses its i | abhavati → abhavat |
| 2208 | Liṅ in six senses: impelling, engaging, permission, respectful urging, deliberation, asking for | bhavān vastraṃ kṣālayet |
| 2209 | Parasmaipada endings of liṅ take yāsuṭ, udātta and ṅit | bhava + t → bhava + yās + t |
| 2210 | The t, th of liṅ take suṭ | bhava + yās + s + t |
| 2211 | In sārvadhātuka liṅ, a non-final s is lost | bhava + yās + s + t → bhava + yā + t |
| 2212 | Yā of a sārvadhātuka after short a → iy | bhava + iy + t → bhavet |
| 2213 | Jhi of liṅ → jus | bhava + yās + us |
| 2214 | Pararūpa for an a-vowel + the u of us; iy for yās comes first | bhindyuḥ; bhaveyuḥ |
| 2215 | The liṅ of benediction is ārdhadhātuka only | bhū + ti (no śap) |
| 2216 | In benediction yāsuṭ is kit | bhū + yās + s + t → bhūyāt |
| 2217 | No guṇa or vṛddhi of ik caused by a git, kit or ṅit affix | bhūyāt, not *\bhoyāt |
| 2218 | Luṅ for the past | bhū → abhūt |
| 2219 | With māṅ, luṅ replaces every lakāra | mā bhavān bhūt |
| 2220 | With māṅ + sma, laṅ (and luṅ) | mā sma bhavat / mā sma bhūt |
| 2221 | Before luṅ, the root takes cli | bhū + cli + t |
| 2222 | Sic replaces cli | bhū + s + t |
| 2223 | Luk of sic after gā, sthā, dā, dhā, pā, bhū before parasmaipada | a-bhū-s-t → a-bhū-t |
| 2224 | No guṇa of bhū, sū before a sārvadhātuka tiṅ | abhūt, not *\abhot |
| 2225 | Īṭ for an apṛkta consonant after a present sic or as | abhūt (no īṭ: sic gone) |
| 2226 | Jus for jhi after sic, an abhyasta, vid | (would reach abhūvan) |
| 2227 | After an elided sic, jus only after ā | a-bhū-an → abhūvan |
| 2228 | No aṭ, āṭ with māṅ | mā bhavān bhūt |
| 2229 | Lṛṅ where a cause of liṅ is present and the action fails | abhaviṣyat |
| 2230 | Upasargas stand only before the root | pra + bhavāni |
| 2231 | N of āni → ṇ after an upasarga r, ṣ; vārttikas on dur, antar | prabhavāṇi; antarbhavāṇi |
| 2232 | N of ni → ṇ optionally, before other roots | praṇibhavati / pranibhavati |
| 2233 | The ṭi of ātmanepada endings of a ṭit lakāra → e | edha + ta → edhate |
| 2234 | A non-pit sārvadhātuka is treated as ṅit | āte counts as ṅit |
| 2235 | The ā of a ṅit affix after short a → iy | edha + āte → edhete |
| 2236 | Thās → se in a ṭit lakāra | edhase; edhe |
| 2237 | Ām in liṭ for an ic-initial root with a heavy vowel, except ṛcch | edh + ām |
| 2238 | Luk of the l-affix after ām | edhām |
| 2239 | Kṛ, bhū, as in liṭ follow the ām form | edhām + kṛ + liṭ |
| 2240 | After-applied kṛ takes ātmanepada as its base does, and only then | edhāṃ + kṛ + ta; indāṃcakāra |
| 2241 | Ta, jha of liṭ → eś, irec | kṛ + e; edhāṃcakrire |
| 2242 | A non-pit liṭ after a non-cluster is kit | kṛ + e (no guṇa) |
| 2243 | No substitute for a vowel before a doubling-causing vowel until doubling is done | kṛ + e → kṛ kṛ e |
| 2244 | The ṛ of the abhyāsa → a (with r) | kṛ kṛ e → kar kṛ e → ka kṛ e |
| 2245 | A ka-series sound or h of the abhyāsa → ca-series | ka kṛ e → cakre; edhāṃcakre |
| 2246 | No iṭ after a one-vowel anudātta root | edhāṃcakṛṣe |
| 2247 | The dh of ṣīdhvam, luṅ, liṭ → ḍh after an iṇ-final aṅga | edhāṃcakṛḍhve |
| 2248 | The initial a of the abhyāsa is lengthened | edhām āsa |
| 2249 | S is lost before an affix beginning with dh | edhitādhve |
| 2250 | The s of tās, as → h before e | edhitāhe |
| 2251 | The e of loṭ → ām | edhatām |
| 2252 | After s, v, the e of loṭ → va, am | edhasva, edhadhvam |
| 2253 | The e of a first person loṭ ending → ai | edhai |
| 2254 | Vowel-initial roots take āṭ in luṅ, laṅ, lṛṅ | aidhata |
| 2255 | Ātmanepada endings of liṅ take sīyuṭ | edheta |
| 2256 | Jha of liṅ → ran | edheran |
| 2257 | Iṭ of liṅ → a | edheya |
| 2258 | In the ātmanepada, jh after a non-a → at; the anudātta roots listed | aidhiṣata; aidhiḍhvam |
| 2259 | In the abhyāsa, a khay after a śar is the consonant kept | spardh → paspardhe |
| 2260 | Before a kit liṭ, a between single consonants → e, and the abhyāsa is lost | dadh dadh e → dedhe |
| 2261 | The same before thal with iṭ; a same-sound substitute does not block | nemitha, dedhāte |
| 2262 | An idit root takes num after its last vowel | skud → skundate, cuskunde |
| 2263 | No e for śas, dad, v-initial roots, or an a made by guṇa | dadade |
| 2264 | Root-initial ṣ → s | ṣvad → svadate, sasvade |
| 2265 | ik before a root's penultimate r/v + consonant is lengthened | urd → ūrdate |
| 2266 | s after iṭ is lost before īṭ (the loss counts as done for merging) | āt-i-s-ī-t → ātīt |
| 2267 | vṛddhi of vad, vraj and consonant-final aṅgas before parasmaipada sic | avādīt, apākṣīt |
| 2268 | No consonant-final vṛddhi before iṭ-initial sic | mā bhavān atīt, acetīt |
| 2269 | After an irit root, aṅ optionally replaces cli | acyutat / acyotīt |
| 2270 | After an upasarga's iṇ, s → ṣ in sunoti … svañj | abhiṣuṇoti, niṣedhati |
| 2271 | s of sad → ṣ after any upasarga but prati | pratisīdati (no ṣ) |
| 2272 | s of the sautra stanbh → ṣ after an upasarga | bāhupratiṣṭambha… |
| 2273 | After ava, s of stanbh → ṣ for support and nearness | avaṣṭabhya, avaṣṭabdhā gauḥ |
| 2274 | After vi and ava, s of svan → ṣ for eating | viṣvaṇati |
| 2275 | After pari, ni, vi, s → ṣ in sev, sita, saya, sivu, saha, suṭ, stu, svañj | pariṣevate, pariṣkaroti |
| 2276 | For the fifteen roots before sita, ṣ even across aṭ | nyaṣedhat |
| 2277 | For the ten from sthā, ṣ across the abhyāsa; abhyāsa ṣ only for them | niṣiṣedha |
| 2278 | No ṣ for sedh meaning "to go" | gaṅgāṃ visedhati |
| 2279 | iṭ optional after svṛ, sū, dhūñ and ūdit roots | siṣedhitha / siṣeddha |
| 2280 | t, th after a jhaṣ → dh (not after dhā) | siṣeddha, seddhā |
| 2281 | s between two jhals is lost | asaiddhām |
| 2282 | Penultimate a → vṛddhi before a ñit/ṇit affix | cakhāda |
| 2283 | Uttama ṇal is optionally ṇit | cakhāda / cakhada |
| 2284 | Light a of a consonant-initial aṅga: optional vṛddhi before iṭ-sic | akhādīt / akhadīt |
| 2285 | n of ni → ṇ after an upasarga r/ṣ before gad and the rest | praṇigadati |
| 2286 | Root-initial ṇ → n | ṇad → nadati |
| 2287 | n of a ṇopadeśa root → ṇ after an upasarga r/ṣ, compound or not | praṇadati |
| 2288 | nuṭ after the lengthened abhyāsa ā of a two-consonant root | ānarda |
| 2289 | Initial ñi, ṭu, ḍu of a root are it-markers | ṭunadi → nandati |
| 2290 | Abhyāsa i/u → iy/uv before an unlike vowel | uvokha; but ūkhatuḥ |
| 2291 | aṅ optional after jṝ, stambhu, mrucu … śvi | amrucat / amrocīt |
| 2292 | aj → vī in the ārdhadhātuka domain (not ghañ, ap) | vivāya, vivyatuḥ |
| 2293 | No iṭ in liṭ only after kṛ, sṛ, bhṛ, vṛ (and stu, dru, sru, śru) | tuṣṭuva; others babhūviva |
| 2294 | No iṭ in thal after a vowel-final root never seṭ in tās | vivetha |
| 2295 | The same for a short-a root never seṭ in tās | papaktha |
| 2296 | Bhāradvāja: only ṛ-final such roots refuse iṭ in thal, hence an option | vivayitha / vivetha, ājitha |
| 2297 | Final ik of an aṅga → vṛddhi before parasmaipada sic | avaiṣīt |
| 2298 | Final vowel long before a y-affix that is not kṛt or sārvadhātuka | kṣīyāt |
| 2299 | No sic-vṛddhi before iṭ for h/m/y-final, kṣaṇ, śvas, jāgṛ, ṇi-final, śvi, edit | akaṭīt |
| 2300 | jhal-initial liṅ and ātmanepada sic are kit after a consonant next to ik | tipsīṣṭa, atipta |
| 2301 | e and loss of the abhyāsa for tṝ, phal, bhaj, trap | trepe |
| 2302 | num for radh and jabh before a vowel | jambhate |
| 2303 | āya after gup, dhūp, vicch, paṇ, pan in their own sense | gup → gopāya |
| 2304 | A form ending in san … ṇiṅ is a root | gopāyati |
| 2305 | āya, īyaṅ, ṇiṅ optional when an ārdhadhātuka is intended | gopāyāṃcakāra / jugopa |
| 2306 | ām in liṭ after kās and after roots of more than one vowel (not in a mantra) | gopāya + ām |
| 2307 | Heading: "before an ārdhadhātuka" | (governs SK 2308) |
| 2308 | Final a of a stem deleted before an ārdhadhātuka | gopāyāṃcakāra |
| 2309 | m of an m-final root → n before m or v | cakṣaṇvahe |
| 2310 | ṇiṅ after kam in its own sense | kāmayate |
| 2311 | ṇi → ay before ām, anta, ālu, āyya, itnu, iṣṇu | kāmayāṃcakre |
| 2312 | caṅ replaces cli after ṇi-final roots and śri, dru, sru in agentive luṅ | kām i a ta |
| 2313 | ṇi is deleted before an ārdhadhātuka not beginning with iṭ; by a vārttika it beats iyaṅ, yaṇ, guṇa, vṛddhi, lengthening | kām i a ta → kām a ta |
| 2314 | Before ṇi + caṅ, the penultimate of the aṅga is shortened | kām a ta → kam a ta |
| 2315 | Before caṅ, the first single-vowelled part of the root is doubled | kam a ta → ka kam a ta |
| 2316 | The abhyāsa is treated as before san (light vowel, then ṇi + caṅ, no ak lost) | ka kam a ta |
| 2317 | Before san, the a of the abhyāsa becomes i | ka kam a ta → ki kam a ta |
| 2318 | Under sanvadbhāva, the light vowel of the abhyāsa is lengthened; vārttika: caṅ after bare kam | acīkamata; acakamata |
| 2319 | Jan, san, khan optionally take ā before a y-initial kit/ṅit affix | sāyāt / sanyāt |
| 2320 | Ṣṭhiv, klam, cam (after āṅ) lengthen before a śit affix | ācāmati, ṣṭhīvati |
| 2321 | Eight roots optionally take śyan | krāmyati / krāmati |
| 2322 | Kram lengthens before a śit affix in the parasmaipada | krāmati, krāmatu |
| 2323 | Snu, kram take iṭ only when nothing causes ātmanepada | akramīt; upakraṃsyate |
| 2324 | Day, ay, ās take ām in liṭ | ayāṃcakre |
| 2325 | After iṭ following iṇ, the dh of ṣīdhvam, luṅ, liṭ optionally becomes ḍh | āyiḍhvam / āyidhvam |
| 2326 | The r of a prefix becomes l before ay | palāyate, plāyate |
| 2327 | Pyāy becomes pī in liṭ and yaṅ | pipye |
| 2328 | After dīp, jan, budh, pūr, tāy, pyāy, ciṇ optionally replaces cli before ta | apyāyi / apyāyiṣṭa |
| 2329 | What follows ciṇ is deleted by luk | a pyāy i ta → apyāyi |
| 2330 | Before parasmaipada sic, the a next to a final l/r takes vṛddhi | mā bhavān ālīt; abhrīt |
| 2331 | After an abhyāsa caused by san or liṭ, the j of ji becomes g | jigāya; jetā |
| 2332 | Dhinv, kṛṇv: v → a, and the affix u | dhinoti |
| 2333 | An affix-u after no conjunct is optionally lost before m/v | dhinvaḥ / dhinuvaḥ; dhinomi |
| 2334 | After such a u, hi is deleted | dhinu; dhinavāma |
| 2335 | ḍh is lost before ḍh; vārttika: liṭ stays kit after ṛ-penultimate roots | gāḍhā; jagṛhe |
| 2336 | Ksa replaces cli after an aniṭ root ending in śal with ik penultimate | aghṛkṣata |
| 2337 | The a of ksa is lost before a vowel-initial ending | aghṛkṣātām |
| 2338 | Akṣ optionally takes śnu | akṣṇoti / akṣati |
| 2339 | Takṣ "to make thin" optionally takes śnu | takṣṇoti / takṣati kāṣṭham |
| 2340 | After iṣ, sah, lubh, ruṣ, riṣ, t-initial ārdhadhātuka takes iṭ optionally | roṣitā / roṣṭā |
| 2341 | Uṣ, vid, jāgṛ take ām in liṭ optionally | oṣāṃcakāra / uvoṣa |
| 2342 | s becomes t before an s-initial ārdhadhātuka | ghatsyati |
| 2343 | Aṅ replaces cli in the parasmaipada after the divādi puṣādi, the dyutādi, ḷdit roots | aghasat |
| 2344 | The abhyāsa of dyut and svāpi takes saṃprasāraṇa | didyute |
| 2345 | Dyut and the rest optionally take parasmaipada in luṅ | adyutat / adyotiṣṭa |
| 2346 | Mid takes guṇa before an affix beginning with an it-ś | medate; mimide |
| 2347 | Vṛt and the rest optionally take parasmaipada before sya, san | vartsyati / vartiṣyate |
| 2348 | After four vṛtādi, no iṭ before s when there is no taṅ or āna | vartsyati; syantsyati |
| 2349 | After anu etc., the s of syand optionally becomes ṣ for a non-living agent | anuṣyandate jalam |
| 2350 | In kṛp, r becomes l and ṛ becomes ḷ | kalpate; cakḷpe |
| 2351 | Kḷp optionally takes parasmaipada in luṭ, before sya and san | kalptāsi / kalpitāse |
| 2352 | After kḷp, no iṭ for tās and s-initial affixes when there is no taṅ or āna | kalptāsi; kalpsyati |
| 2353 | In liṭ, the abhyāsa of vyath takes saṃprasāraṇa | vivyathe |
| 2354 | Seven phaṇādi optionally take e and lose the abhyāsa | pheṇatuḥ / paphaṇatuḥ; rejatuḥ |
| 2355 | Pat takes the augment pum before aṅ | apaptat |
| 2356 | Jṝ, bhram, tras optionally take e and lose the abhyāsa | bhrematuḥ / babhramatuḥ |
| 2357 | After loss of ḍh, the a of sah, vah becomes o | soḍhā |
| 2358 | No ṣ for the s of sah in the form soḍh | parisoḍhā |
| 2359 | After pari, ni, vi, the s of siv etc. optionally becomes ṣ, even across aṭ | paryasahata / paryaṣahata |
| 2360 | Before a śit affix, pā … sad become piba … sīda | sīdati, tiṣṭhati |
| 2361 | In liṭ, no ṣ for the s of sad after the abhyāsa | niṣasāda |
| 2362 | Śad takes ātmanepada when a śit affix is to follow | śīyate; śaśāda |
| 2363 | Gam, han, jan, khan, ghas lose the penultimate before a vowel-initial kit/ṅit, not before aṅ | cakhnatuḥ; agamat |
| 2364 | The u of guh becomes ū where guṇa would come before a vowel | gūhati; goḍhā |
| 2365 | For duh, dih, lih, guh, ksa is optionally lost before a dental taṅ ending | agūḍha / aghukṣata |
| 2366 | After ṛ-final roots and han, sya takes iṭ | bhariṣyati |
| 2367 | Final ṛ becomes ri before śa, yak, a y-initial ārdhadhātuka liṅ | bhriyāt; kriyate |
| 2368 | After ṛ, a jhal-initial liṅ and a jhal-initial sic before taṅ are kit | bhṛṣīṣṭa; but abhārṣīt |
| 2369 | After a short-vowel stem, sic is deleted before a jhal | abhṛta; but abhṛṣātām |
| 2370 | A root ending in e, o, ai, au in its taught form takes ā, not before a śit | dhe → dhā; but dhayati |
| 2371 | After an ā-final stem, ṇal becomes au | dadhau, jaglau |
| 2372 | ā is lost before a vowel-initial kit/ṅit ārdhadhātuka and before iṭ; after doubling | dadhatuḥ, dadhitha |
| 2373 | Roots of the form dā and dhā are called ghu, except dāp, daip | dheṭ is ghu |
| 2374 | Ghu, mā, sthā etc. take e before a kit ārdhadhātuka liṅ | dheyāt, stheyāt |
| 2375 | After dheṭ and śvi, caṅ optionally replaces cli | adadhat, aśiśviyat |
| 2376 | After ghrā, dheṭ, śā, chā, sā, luk of sic is optional in parasmaipada | adhāt, adhuḥ |
| 2377 | yam, ram, nam and ā-final roots take sak, and their sic takes iṭ, in parasmaipada | adhāsīt, ayaṃsīt |
| 2378 | Other cluster-initial ā-final roots take e optionally before a kit ārdhadhātuka liṅ | glāyāt / gleyāt |
| 2379 | A cluster-initial ṛ-final stem takes guṇa in liṭ, even before ṇal | jahvāra, jahvaratuḥ |
| 2380 | ṛ and cluster-initial ṛ-final stems take guṇa before yak and a y-initial ārdhadhātuka liṅ | hvaryāt, aryāt |
| 2381 | No iṭ before a kit (or git) after śri and one-vowelled uk-final stems | śritaḥ; but sasvariva |
| 2382 | aṅ for cli after the juhotyādi sṛ, ṛ and after śās | bhvādi asārṣīt keeps sic |
| 2383 | Guṇa in liṭ for ṛcch, the root ṛ and ṝ-final roots | āra, āratuḥ, teratuḥ |
| 2384 | After ad, ṛ, vyeñ, thal always takes iṭ | āritha, vivyayitha |
| 2385 | After stu, su, dhūñ, sic takes iṭ in parasmaipada | asāvīt (or asauṣīt) |
| 2386 | śru becomes śṛ and takes śnu for śap | śṛṇoti, śṛṇutaḥ |
| 2387 | The u of hu and of a śnu-stem becomes v before a vowel-initial sārvadhātuka | śṛṇvanti |
| 2388 | deṅ becomes digi in liṭ, with no doubling | digye |
| 2389 | sthā and ghu take i, and sic is kit, before taṅ | adita, asthita |
| 2390 | ṝ-final roots take i; guṇa and vṛddhi win by a vārttika | tarati, tīryāt |
| 2391 | After vṛ and ṝ-final roots, iṭ is optionally long, not in liṭ | tarītā / taritā; teritha |
| 2392 | No long iṭ before a parasmaipada sic | atāriṣṭām |
| 2393 | san after gup, tij, kit | jugupsate, cikitsati |
| 2394 | san after mān, badh, dān, śān, with long i in the abhyāsa; vārttikas fix the senses | mīmāṃsate, bībhatsate |
| 2395 | In a form ending in san or yaṅ, the first one-vowelled part is doubled | jugupsate, titikṣate |
| 2396 | daṃś, sañj, svañj lose their n before śap | svajate, daśati |
| 2397 | rañj too loses its n before śap | rajati; pariṣasvaje |
| 2398 | After vi, s of skand optionally becomes ṣ, not before a niṣṭhā | viṣkantā / viskantā; viskannaḥ |
| 2399 | After pari, the same option, even before a niṣṭhā | pariṣkandati / pariskandati |
| 2400 | iṣ, gam, yam take ch before a śit | gacchati, yacchati |
| 2401 | After gam, an s-initial ārdhadhātuka takes iṭ before parasmaipada endings | gamiṣyati |
| 2402 | Anudātta roots with penultimate ṛ optionally take am before a jhal-initial non-kit affix | sraptā / sarptā |
| 2403 | The s of nis becomes ṣ before tap, except for repetition | niṣṭapati |
| 2404 | After sṛj and dṛś, thal optionally takes iṭ | dadarśitha / dadraṣṭha |
| 2405 | sṛj and dṛś take am before a jhal-initial non-kit affix | draṣṭā, dadraṣṭha |
| 2406 | Guṇa before aṅ for ṛ-final stems and dṛś | adarśat |
| 2407 | After dṛś, cli does not become ksa; a vārttika gives optional sic for kṛṣ etc. | adrākṣīt; akrākṣīt / akṛkṣat |
| 2408 | In liṭ the abhyāsa of the yajādi and grahādi takes saṃprasāraṇa | iyāja |
| 2409 | vac, svap and the yajādi take saṃprasāraṇa before a kit | ījatuḥ, ijyāt, ūpe |
| 2410 | The s of śās, vas, ghas becomes ṣ after iṇ or k | ūṣatuḥ |
| 2411 | veñ is optionally replaced by vay in liṭ | uvāya |
| 2412 | grah, jyā, vay etc. take saṃprasāraṇa before a ṅit or kit | gṛhṇāti, jināti |
| 2413 | In liṭ the y of vay takes no saṃprasāraṇa | ūyatuḥ |
| 2414 | In a kit liṭ that y optionally becomes v | ūvatuḥ / ūyatuḥ |
| 2415 | veñ (without vay) takes no saṃprasāraṇa in liṭ | vavau, vavatuḥ |
| 2416 | vyeñ takes no ā in liṭ | vivyāya, vivyatuḥ |
| 2417 | hveñ takes saṃprasāraṇa before it is doubled | juhāva, juhuvatuḥ |
| 2418 | aṅ for cli after lip, sic, hvā | alipat, ahvat |
| 2419 | Before ātmanepada that aṅ is optional | ahvata / ahvāsta |
| 2420 | śvi optionally takes saṃprasāraṇa in liṭ and before yaṅ; no abhyāsa saṃprasāraṇa by a vārttika | śuśāva / śiśvāya |
| 2421 | The i of śvi becomes a before aṅ; the bhvādi ends | aśvat, aśiśviyat, aśvayīt |
| 2422 | īyaṅ after the sautra root ṛt, in its own sense | ṛtīyate; ānarta |