Rule. For the word ahan, ru replaces the final n at the end of a pada. Ru is the substitute r with a marker u; its later sandhi you know already.
Why does the sūtra need saying at all, and why is its single word ahan read twice? Because both this ru-substitution and the related r-substitution lie in the Tripādī and are therefore asiddha — treated as not yet having happened — for the earlier rule that drops a final n (na-lopa). Left alone, na-lopa would strike first and there would be no n left to replace. The first reading of ahan settles by fiat (nipātana) that na-lopa does not apply to ahan; the second reading enjoins ru.
Example.
ahan standing as a pada → n → ru → the ru goes to visarga at pause → ahaḥ.
ahan + bhyām → the stem is a pada before bhyām → n → ru → ru → o before the voiced bh → ahobhyām; likewise ahobhiḥ.
The rule also reaches a word that ends in ahan (a tadanta). Take dīrghāhan, "that in which the days are long", i.e. summer (nidāgha):
dīrghāhan + su → su is dropped by the rule hal-ṅyādi-lopa (a consonant-stem loses its su) → the dropped su still counts as present by pratyaya-lakṣaṇa (a dropped affix still conditions the rules it would have conditioned) → therefore the r-substitution, which is worded "when no sup follows" (asupi), is barred → only ru applies → but ru is asiddha, so for the lengthening rule the stem still ends in n → the vowel before that n is lengthened, as in rājā → dīrghāhā.
Vocative: he dīrghāho nidāgha "O long-dayed summer!" (no lengthening in the vocative; the ru gives o before n).
dīrghāhānau (du.), dīrghāhānaḥ (pl.), dīrghāhnā (instr. sg.), dīrghāhobhyām (instr. du.).
Applies when / blocked by. Only at the end of a pada, and only for ahan or a word ending in it. The competing r-substitution is barred whenever a case-ending, even a dropped one, follows.
Watch out. Two Tripādī facts do the work here. Ru is asiddha for na-lopa, so na-lopa would have won without the nipātana; and ru is asiddha for the lengthening rule, so dīrghāhā is lengthened as though its n were still there.
The survey: neuter in-stems, asṛj, ūrj
Bhaṭṭoji now declines the remaining consonant-stems in the neuter by cross-reference. In the neuter singular the endings su and am are dropped, so a stem in -in loses its n at pada-end; in the dual śī replaces the dual ending au; in the plural śi is a strong ending, so the vowel before the final n is lengthened, as in rājan → rājā-. No num is added: an n-stem is neither ugit nor jhal-final.
daṇḍin + śi → śi is a strong ending → the i before the final n is lengthened → daṇḍīni; no num.
| stem | nom./acc. sg. | du. | pl. |
|---|---|---|---|
| daṇḍin "having a staff" | daṇḍi | daṇḍinī | daṇḍīni |
| sragvin "garlanded" | sragvi | sragviṇī | sragvīṇi |
| vāgmin "eloquent" | vāgmi | vāgminī | vāgmīni |
Compounds ending in -han, -pūṣan and -aryaman form their neuter plural in the same way, by lengthening alone, without num: bahuvṛtrahāṇi, bahupūṣāṇi, bahvaryamāṇi.
asṛj "blood". Its final j becomes a velar at the end of a pada, because the affix kvin is enjoined in the formation of sṛj (chapter 11). Bhaṭṭoji adds three remarks. The velar does not appear in viśvasṛj and the like — the Mahābhāṣya uses the form rajju-sṛḍbhyām; or else, by SK 294 (vraśca…), sṛj and yaj take ṣ at pada-end, which is an exception to the velar; and in srak and ṛtvik the velar is by fiat alone. The word asṛk itself, he says, is to be taken as formed from the root as (asyati) with the uṇādi affix ṛc.
asṛj → pada-end → asṛk (at pause) / asṛg (before a voiced sound).
asṛj + śī → asṛjī.
asṛj + śi → śi is a strong ending, the stem ends in a jhal → num → asṛñji.
Or, with the substitute stem asan of SK 228: asāni (pl.), asnā (instr. sg.).
asṛjā (instr. sg.), asṛgbhyām, asabhyām (instr. du.), and so on.
ūrj "vigour": ūrk / ūrg; ūrjī (du.); ūrñji (pl., a cluster of n, r and j).
Two vārttikas concern the compound bahūrj, used of families:
- bahūrji num-pratiṣedhaḥ — in bahūrji the num is prohibited.
- antyāt pūrvo vā num — optionally, num is inserted before the final sound.
So the neuter plural is bahūrji, or optionally bahūrñji (kulāni, "families").
The survey: the pronouns of the tyad list, and bebhid
| stem | nom./acc. sg. | du. | pl. |
|---|---|---|---|
| tyad | tyat / tyad | tye | tyāni |
| tad | tat / tad | te | tāni |
| yad | yat / yad | ye | yāni |
| etad | etat / etad | ete | etāni |
When etad refers back to something just mentioned (anvādeśa), the neuter is enat.
bebhid is made by adding the affix kvip to bobhidyati, the intensive of bhid "split": bebhit / bebhid; bebhidī (du.); bebhidi (pl., of Brāhmaṇa families that keep splitting). Note the plural has no num. The final a of the intensive stem was dropped before kvip (al-lopa), and that dropped a is sthānivat — it counts as still present for rules applying to what precedes it. So the stem does not end in a jhal, and the num for jhal-stems fails. Nor can the stem claim the num for vowel-stems: a dropped sound is not treated as present in an operation that concerns that very sound (sva-vidhi). This same reasoning returns for pipaṭhiṣi in Section D.
The survey: gavāk and its hundred and nine forms
gavāk "going toward a cow" is analysed gām añcati, with the affix kvin (SK 373). Its neuter forms multiply for four reasons, which Bhaṭṭoji's two counting-verses put at one hundred and nine in all:
- añc means "motion" (gati) or "worship" (pūjā). In the motion sense the n of añc is dropped (chapter 11); in the worship sense it is kept, and at pada-end becomes the velar nasal ṅ.
- go before a vowel takes the avaṅ substitute (SK 88): gav-.
- Or the final o of go is left uncombined (SK 87, prakṛti-bhāva): go-a-.
- Or the a after o is absorbed into it (pūrva-rūpa): go'-.
gām añcati → kvin → motion sense: n dropped → avaṅ → gavāk / gavāg; uncombined → go-ak / go-ag; pūrva-rūpa → go'k / go'g. Worship sense: n kept, ṅ at pada-end → gavāṅ, go-aṅ, go'ṅ. Nine forms in the nominative singular; the same nine in the accusative singular.
Dual: au → śī → the stem is now a bha → the a of añc is dropped (SK 416) → gocī; in the worship sense gavāñcī, go-añcī, go'ñcī.
Plural: jas / śas → śi (SK 312) → śi is a strong ending → num → gavāñci, go-añci, go'ñci. Only three forms here, for both senses.
Instrumental singular: gocā; gavāñcā, go-añcā, go'ñcā.
Before bhyām: gavāgbhyām, go-agbhyām, go'gbhyām, gavāṅbhyām, go-aṅbhyām, and so on.
Locative plural: the ṅ-final forms optionally take the augment kuk (SK 130) → gavāṅkṣu, gavāṅṣu, gavākṣu, go-akṣu, gokṣu.
The count: nine forms each in su, am and the locative plural; six in each of the six endings beginning with bh; three in the nominative and accusative plural; four in each of the remaining ten cases. That is 27 + 36 + 6 + 40 = 109. Bhaṭṭoji adds two cautions. One might fear that the optional change of k to kh (cayo dvitīyāḥ) adds six forms; it does not, because that cartva is asiddha. But on the kuk-alternative, where no jaśtva applies, the kh-alternative does add three forms. Beyond that, the optional doubling and nasalization of consonants multiply the forms further, as they do for aśvākṣi (SK 527); the exact sense of this closing verse scholars discuss.
The survey: tiryañc, yakṛt, śakṛt
tiryañc "going across": tiryak, tiraścī, tiryañci; in the worship sense tiryaṅ, tiryañcī, tiryañci.
yakṛt "liver": yakṛt, yakṛtī, yakṛnti; or with the substitute stem yakan of SK 228: yakāni (pl.), yaknā (instr. sg.), beside yakṛtā.
śakṛt "dung": śakṛt, śakṛtī, śakṛnti; śakāni, śaknā, śakṛtā.
Then dadat "giving": dadat, dadatī … — and here the participle rules begin.
For [the word] ahan, ru [is substituted for its final n] at pada-end. [Thus:] ahobhyām; ahobhiḥ. Here, in "ahaḥ," "ahobhyām," etc., since the ratva and rutva (r- and ru-substitutions) are asiddha (regarded as not-yet-effected), n-elision (na-lopa) would [otherwise] obtain; [therefore,] by repeating "ahan" [in the sūtra text], the non-application of na-lopa is settled by fiat (nipātana), and by the second [occurrence of "ahan"] ru is to be enjoined. Even for [a compound] ending in that (tadanta), there is rutva and ratva. "That [summer] in which the days are long" — [the compound] dīrghāhā (= summer, nidāgha). Here [in dīrghāhā, nom. sg.], when su is elided by "hal-ṅyādi[-lopa]," although by pratyaya-lakṣaṇa [the elided su would still condition], because of the prohibition "asupi" ratva does not apply, [so only] ru [applies]; and because that [ru] is asiddha, the upadhā-lengthening conditioned on an n-ending [stem] [applies]. But in the vocative (saṃbuddhi): "he dīrghāho nidāgha" (O long-summered one!). dīrghāhānau (nom./acc. du.); dīrghāhānaḥ (nom. pl.); dīrghāhnā (instr. sg.); dīrghāhobhyām (instr. du.).
daṇḍi; daṇḍinī; daṇḍīni [neut. of daṇḍin]. sragvi; sragviṇī; sragvīṇi [neut. of sragvin]. vāgmi; vāgminī; vāgmīni [neut. of vāgmin]. bahuvṛtrahāṇi; bahupūṣāṇi; bahvaryamāṇi [neut. pl. of compounds ending in -han, -pūṣan, -aryaman].
For asṛj, [there is] kutva (guttural substitution) at pada-end, because kvin is enjoined [in the formation] of sṛj. But not in viśvasṛj etc. — because in the sūtra "sṛji-dṛśoḥ" (SK 2405) the Bhāṣya uses the form "rajju-sṛḍbhyām." Or else: by the sūtra "vraśca…" (SK 294), for sṛj and yaj there is ṣatva at pada-end, [which is] the exception (apavāda) to kutva. But for the words srak (sraj) and ṛtvik, the kutva is by fiat (nipātana) alone. The word asṛk ("blood"), however, is to be understood [as formed] from the root "as" (asyati) with the uṇādi affix ṛc. asṛk (nom. sg. neut., at pause); asṛg (before a voiced sound); asṛjī (du.); asṛñji (pl., with num); or, by "pad-dan…" (SK 228), [the substitute stem] asan: asāni (pl.); asṛjā (instr. sg.); asnā (instr. sg. of the asan-stem); asṛgbhyām; asabhyām, etc. ūrk; ūrg; ūrjī (du.); ūrñji (pl.) — [here] a conjunct of n, r, and j.
(vārttika) "bahūrji num-pratiṣedhaḥ" — In bahūrj[i], num is prohibited (pratiṣedha).
(This is a vārttika.)
(vārttika) "antyāt pūrvo vā num" — Optionally (vā), num [is inserted] before the final [phoneme].
(This is a vārttika.) [So:] bahūrji, or optionally bahūrñji — "the families" (kulāni). tyat; tyad; tye; tyāni [neut. of tyad]. tat; tad; te; tāni [neut. of tad]. yat; yad; ye; yāni [neut. of yad]. etat; etad; ete; etāni [neut. of etad]. But in a re-mention (anvādeśa): enat. kvip [added] to bobhidyati (the intensive of bhid): bebhit; bebhid; bebhidī (du.). Before śi (śau), because the al-lopa (elision of the final a [of bebhidya]) is sthānivat, [the stem being thereby] non-jhal-ending (ajhalanta), there is no num; and num conditioned on an ac(vowel)-ending [stem] [also] does not [apply], because there is no sthānivadbhāva in an operation concerning [the substituend] itself (sva-vidhi). bebhidi — "if the Brāhmaṇa-families [are] cleaving" (chidi).
[Verse 1:] The forms of the word gavāk in the neuter, owing to the difference between [the senses of] "arcā" (worship) and "gati" (motion), together with [the variants from] non-saṃdhi, avaṅ, and pūrva-rūpa, are held to be one hundred and nine (nava-adhika-śatam). [Verse 2:] Consider (vibhāvaya) [the count] thus: in su-am-sup — nine [each]; in the bh-initial set of six [endings] — six; in jas/śas — three; and four in each of the remaining ten [cases].
To wit (tathā hi): in the analysis "gām añcati" [that which goes toward the cow/ earth], by "ṛtvik" (SK 373) etc. [there is] kvin. In the sense of "motion" (gati), [there is] na-lopa (elision of the n of añc). By "avaṅ sphoṭāyanasya" (SK 88), the avaṅ [substitute]. gavāk; gavāg. By "sarvatra vibhāṣā" (SK 87), in the prakṛti-bhāva (retention of the un-combined form): go-ak; go-ag. With pūrva-rūpa: go'k; go'g. In the sense of "worship" (pūjā), by kutva of the n [it becomes] ṅ: gavāṅ; go-aṅ; go'ṅ. In the am (acc. sg.) too, these very [same] nine [forms occur]. śī [replaces] auṅ [the dual affix]. Because [the stem] is [now] bha, by "acaḥ" (SK 416) [there is] al-lopa: gocī. But in the pūjā sense: gavāñcī; go-añcī; go'ñcī. By "jaśśasoḥ śiḥ" (SK 312), jas/śas → śi; because śi is a sarvanāmasthāna, num: gavāñci; go-añci; go'ñci. In the gati and pūjana [senses], [there are] only three [in jas/śas]. gocā; gavāñcā; go-añcā; go'ñcā. gavāgbhyām; go-agbhyām; go'gbhyām; gavāṅbhyām; go-aṅbhyām; and so on.
But in the sup (loc. pl.), for the ṅ-ending [pūjā forms], optionally, by "ṅṇoḥ kuk" (SK 130), the kuk-augment: gavāṅkṣu; gavāṅṣu; gavākṣu; go-akṣu; gokṣu. And one should not suspect here [an increase in the count]: on the alternative [where], by "cayo dvitīyāḥ," the k-kāra [becomes] kh-kāra, [one might fear] an excess of six forms — [but there is none,] because the cartva is asiddha. On the kuk-alternative, however, since that [kuk] is asiddha, when jaśtva is absent, on the alternative of the second-substitution [kh by "cayo dvitīyāḥ"] three forms do indeed increase [the count]. [Verse:] It is to be inferred (ūhyam) that, owing to the optionality of doubling (dvirvacana) and nasalization (anunāsika), the forms of these become "aśvākṣi-bhūta" [i.e. five hundred and twenty-seven] — so [say] the wise.
tiryak; tiraścī; tiryañci. But in the pūjā sense: tiryaṅ; tiryañcī; tiryañci. yakṛt; yakṛtī; yakṛnti; or, by "pad-dan" (SK 228), [the substitute stem] yakan: yakāni; yaknā; yakṛtā. śakṛt; śakṛtī; śakṛnti; śakāni; śaknā; śakṛtā. dadat; dadatī…
- translator aśvākṣi-bhūta in the closing verse is a word-numeral (bhūta-saṃkhyā): aśva = 7, akṣi = 2, bhūta = 5, read from the right, giving 527, as TB states (sapta-viṃśaty-adhikā pañcaśatī) and as BM's case-by-case tally (on sk.442.b1) totals. The source tags the word as a cross-reference to SK 527, a mis-tagging of the numeral; the reference is dropped from the English. The count-scholasticism of the two gavāk verses is otherwise translated without independently verifying the enumeration (one hundred and nine).
- textual The paradigm forms ūrk / ūrg / ūrjī / ūrñji show inconsistent vowel length in the source (e.g. "urk" with short u, "ūrji"); regularized here to the ūrj-stem. The plural is transmitted as "ūnrji" (an anomalous cluster order) and read as ūrñji, which the following gloss "a conjunct of n, r, and j" describes.
- textual The Bālamanoramā block formerly filed here as sk.443.b1 (on hali ca) was re-filed to sk.354.b1 on 2026-10-05; its translation moved with it.
- textual The Tattvabodhinī on this sūtra is sk.443.t1, re-filed 2026-10-05 from sk.441.t1 (p12, under yaḥ sau); the TB notes in this unit are made from it. The block runs on at both ends: its opening (pratīkas datvam, pūrvapadasyeva, uttarapadatve ca, dadhisecau, anvādeśa, brahmaṇī, ro 'supi / aharbhāti) is on the opening of the halanta-napuṃsaka section (svanaḍut … ahaḥ), whose kaumudī text is absent from our units (the BM on it stands in sk.442.b1); its last pratīka, śatā, belongs to SK 444 (vā napuṃsakasya).
- commentary TB on 'nalopābhāvaṃ nipātya': even if a second n is read into 'ahan' (ru for the n-final word ahan), the rutva and ratva still have the n as their sthānin. On 'tadantasyāpi': because the rule stands in the pada-adhikāra; reaching the compound-final ahan as a pada through the internal vibhakti by pratyaya-lakṣaṇa is not available, 'uttarapadatve cāpadādividhau' prohibiting it. Followed for 'even for [a compound] ending in that (tadanta)'.
- commentary TB on 'in the sūtra sṛji-dṛśoḥ': an upalakṣaṇa, the Bhāṣya using rajjusṛḍbhyām under 'mṛjer vṛddhiḥ' too. TB explains why rajjusṛḍbhyām takes no am-augment by 'sṛji-dṛśoḥ…' before the jhal-initial, non-kit bhyām: an operation stated for a dhātu by its own form applies only before a suffix introduced after that dhātu, as the Bhāṣya on 'mṛjer vṛddhiḥ' makes clear. On 'kutvāpavāda': because the ṣatva sūtra names sṛj, mṛj and yaj specifically. Followed; the English already has ṣatva as the apavāda to kutva.
- commentary TB on asṛk: the derivation from as with the uṇādi affix ṛc is clear in Kṣīrasvāmin's work; though this analysis changes the accent, accent is disregarded in ordinary usage (loka), and in the Veda the desired accent follows by bāhulaka. Recorded; nothing changed.
- commentary TB: 'antyāt pūrvo vā num' is likewise a stated (vācanika) provision; the Bhāṣya's vārttika reads 'bahūrji pratiṣedhaḥ, antyāt pūrvaṃ num eke'. The Bhāṣya rejects the first vārttika by taking 'napuṃsakasya jhalacaḥ' as num for a stem ending in a jhal that follows an ac; on that view ūrj gets no num (against the Kaumudī's ūrñji), nor do kvip-derivatives of urd and garh. BM (sk.442.b1) likewise sets aside the Kaumudī's 'nara-jānāṃ saṃyogaḥ' as contrary to the Bhāṣya. Recorded (§16); the Kaumudī's forms are translated as they stand.
- commentary TB on anvādeśa (said of the enat of idam in the section opening, bearing equally on enat here): enat is enjoined only for am, not for auṭ, śas etc., there being no difference in result; pratyaya-lakṣaṇa operates despite the luk of am by 'svamor napuṃsakāt', by force of the statement. In truth enat should be stated in 'dvitīyāṭaussu…' itself; the t pronounced in it secures enat despite the luk, and it does not reach the compound etac-chritam, whose luk is bahiraṅga (so the Manoramā). Followed for 'in a re-mention: enat' as the acc. sg.
- commentary TB on 'al-lopasya sthānivattvāt' (bebhid): the maxim 'kvau luptaṃ na sthānivat' is not invoked here, being of limited scope; the general one is 'kvau vidhiṃ prati na sthānivat'. TB's pratīka begins at al-lopasya, and BM (sk.442.b1, 'śāv iti': the śi of jas/śas, where a jhal-conditioned num is apprehended) shows that the transmitted 'śāv allopasya' is 'śau, al-lopasya…'. English corrected from 'the al-lopa [caused] by the śit [kvip]' to 'Before śi (śau), because the al-lopa … is sthānivat'.
- commentary TB on the gavāk verses: earlier paṇḍitas gave the two verses as the answer to an old challenge-verse posed in a king's assembly (nine forms each in su and am, six in bhyām/bhis/bhyas, nine in sup, three each in jas and śas, four in the remaining endings, six months allowed for the answer); 'arcā-gati-bhedataḥ' has tasi by ādyāditva; 'each' (prati-ekam) is to be supplied in 'svam-sup-su' and with the six, three and four. The 109 leaves out the vārttika 'cayo dvitīyāḥ…'; TB lists the further forms covered by 'ityādi', and answers that forms identical across bhyām, bhyas, ṅasi/ṅas and os are counted separately by difference of meaning. Followed.
- commentary TB on 'ṣaṇṇām ādhikyam…': the feared excess of six is the optional kh for the k among the nine forms; 'cartvasya asiddhatvāt' means that from the standpoint of the śāstra there is no k there, only g; 'tasya asiddhatvāt' = 'because the kuk is asiddha'; 'vardhanta eva': with three more, sup has twelve forms and the total is 112, not 109. Followed: the bracket in the kuk sentence corrected from [cartva] to [kuk]. TB then works out the 527 of aśvākṣi-bhūtāni case by case from optional doubling (citing 'yaṇo mayaḥ', 'khayaḥ śaraḥ') and optional nasalization by 'aṇo 'pragṛhyasya' (none in the auṅ forms, which are pragṛhya, nor before e, which is not aṇ). Followed in giving the number.
- commentary TB on tiraścī: al-lopa by 'acaḥ' because the stem is bha, and no tiri-substitution because of 'alope'; on dadat: already derived in the masculine. Agrees with BM; nothing changed.