Skip to content

Chapter 6: An Outline of Sanskrit Grammar: Sandhi, Compounds, Case and Verb

arthapravṛttitattvānāṃ śabdā eva nibandhanam
tattvāvabodhaḥ śabdānāṃ nāsti vyākaraṇād ṛte

Words alone are the means by which the truth about things and their uses is fixed; and without grammar there is no true understanding of words.

— Bhartṛhari, Vākyapadīya 1.13

6.1 Introduction

Chapter 5 described the Aṣṭādhyāyī as a system of rules. This chapter is its companion: a reference outline of the Sanskrit that those rules describe, for readers who want to see the language itself before following the grammarians’ debates in the chapters that come next. It covers the combination of sounds (sandhi), the formation of compounds (samāsa), the case system, the derivation of words by suffixes, and the verb. Pāṇini’s rule numbers are given where they help to show how the traditional grammar treats each topic. For fuller modern accounts, see the standard reference grammars and Cardona (1997); for the Pāṇinian treatment of individual rules, Katre (1987) and Sharma (1987–2003).

6.2 Sandhi

Sandhi (“joining”) is the set of sound changes that take place where elements meet: within a word, between the members of a compound, and between words in a sentence. In written Sanskrit external sandhi is normally applied, so that a sentence may appear as one continuous string. This is one reason why reading Sanskrit, and processing it by computer (Chapter 14), requires a sound knowledge of sandhi.

6.2.1 Vowel sandhi

The principal rules are set out in Table 6.1; Figure 6.1 arranges them as a sequence of decisions.

Table 6.1. Principal rules of vowel sandhi
Meeting vowels Result Rule and example
similar vowels (a/ā + a/ā, i/ī + i/ī …) long vowel 6.1.101; deva + ālaya \(\rightarrow\) devālaya
a/ā + i/ī e 6.1.87 (guṇa); deva + indra \(\rightarrow\) devendra
a/ā + u/ū o 6.1.87; sūrya + udaya \(\rightarrow\) sūryodaya
a/ā + ṛ ar 6.1.87; mahā + ṛṣi \(\rightarrow\) maharṣi
a/ā + e/ai ai 6.1.88 (vṛddhi); sadā + eva \(\rightarrow\) sadaiva
a/ā + o/au au 6.1.88; mahā + oṣadhi \(\rightarrow\) mahauṣadhi
i/ī, u/ū, ṛ + dissimilar vowel y, v, r + vowel 6.1.77; iti + ādi \(\rightarrow\) ityādi
word-final e, o + a e, o; a elided 6.1.109; vane + atra \(\rightarrow\) vane ’tra
Simplified decision chart for the main rules of vowel sandhi
Figure 6.1. Simplified decision chart for the main rules of vowel sandhi

6.2.2 Consonant sandhi

Table 6.2 gives some of the commonest rules for consonants.

Table 6.2. Some rules of consonant sandhi
Process Description Example
Voicing A word-final stop becomes voiced before a voiced sound vāk + devī \(\rightarrow\) vāgdevī
Palatal assimilation Final t becomes c or j before palatals tat + ca \(\rightarrow\) tac ca; tat + jayati \(\rightarrow\) taj jayati
Nasal assimilation A final stop may become the nasal of its class before a nasal tat + na \(\rightarrow\) tan na; jagat + nātha \(\rightarrow\) jagannātha
Sibilant after a stop After a stop, initial ś may become ch tat + śrutvā \(\rightarrow\) tac chrutvā
Anusvāra Final m becomes anusvāra before a consonant tam + karoti \(\rightarrow\) taṃ karoti
Retroflexion of n n becomes ṇ after r, ṛ, ṣ within a word, if nothing but vowels, gutturals, labials, h, y, v or anusvāra intervene rāmeṇa, karaṇa (8.4.1–2)
Retroflexion of s s becomes ṣ after i, u, ṛ, e, o, k, r agni + su \(\rightarrow\) agniṣu (8.3.59)

6.2.3 Sandhi of visarga

The visarga (-ḥ) changes according to the sound that follows (Table 6.3).

Table 6.3. Main rules for final -ḥ (visarga)
Before Change Example
c, ch ḥ \(\rightarrow\) ś rāmaḥ + ca \(\rightarrow\) rāmaś ca
ṭ, ṭh ḥ \(\rightarrow\) ṣ rāmaḥ + ṭīkām \(\rightarrow\) rāmaṣ ṭīkām
t, th ḥ \(\rightarrow\) s rāmaḥ + tatra \(\rightarrow\) rāmas tatra
k, kh, p, ph, sibilants ḥ remains rāmaḥ karoti
voiced consonant (after a) -aḥ \(\rightarrow\) -o rāmaḥ + gacchati \(\rightarrow\) rāmo gacchati
a (after a) -aḥ \(\rightarrow\) -o, a elided rāmaḥ + atra \(\rightarrow\) rāmo ’tra
other vowels (after a) -aḥ \(\rightarrow\) -a rāmaḥ + iti \(\rightarrow\) rāma iti
voiced sound (after i, u etc.) ḥ \(\rightarrow\) r muniḥ + gacchati \(\rightarrow\) munir gacchati

6.3 Compounds (samāsa)

Sanskrit forms compounds very freely: two or more stems are joined, and only the last carries a case ending. Pāṇini treats compounds in Adhyāya 2.1–2.2; the tradition distinguishes four main types (Table 6.4).

Table 6.4. The four main types of compound
Type Character Example
avyayībhāva Adverbial; first member usually an indeclinable yathāśakti “according to one’s ability”
tatpuruṣa Determinative; the last member is qualified by the first rājapuruṣa “the king’s man”
dvandva Copulative; members coordinated rāmalakṣmaṇau “Rāma and Lakṣmaṇa”
bahuvrīhi Possessive; the compound describes something outside itself pītāmbara “(one) whose garment is yellow”, i.e. Kṛṣṇa

The name bahuvrīhi is itself an example: “(one) who has much rice”, i.e. a wealthy man. The same two words can form compounds of different types with different meanings: nīlotpala as a tatpuruṣa (more precisely a karmadhāraya) means “blue lotus”; as a bahuvrīhi it would mean “(a pond) that has blue lotuses”. In the Vedic language the accent distinguished such pairs, as the story of indraśatru shows (Chapter 2).

6.3.1 Kinds of tatpuruṣa

Tatpuruṣa compounds are named after the case relation of the first member to the second (Table 6.5).

Table 6.5. Tatpuruṣa compounds named after the case relation of the first member, with traditional examples
Relation Example
accusative kṛṣṇaśrita “resorting to Kṛṣṇa”
instrumental śaṅkulākhaṇḍa “cut with a knife”
dative yūpadāru “wood for a sacrificial post”
ablative corabhaya “fear of thieves”
genitive rājapuruṣa “king’s man”
locative akṣaśauṇḍa “skilled at dice”
Karmadhāraya
a tatpuruṣa whose members refer to the same thing: nīlotpala “blue lotus”, mahārāja “great king”.
Dvigu
a karmadhāraya whose first member is a numeral: triloka or trilokī “the three worlds”, pañcavaṭī “(the place of) five banyan trees”.
Nañ-tatpuruṣa
a compound with the negative prefix a(n)-: abrāhmaṇa “non-Brahmin”, adharma.
Upapada
a compound whose second member is a verbal derivative that occurs only in composition: kumbhakāra “potter”, literally “pot-maker”.
Aluk
a compound in which the first member keeps its case ending: parasmaipada (“word for another”), with the dative parasmai.

Pāṇini also lists irregular compounds in a group beginning with mayūravyaṃsaka (2.1.72), which are accepted as they occur in usage.

6.3.2 Long compounds

Classical prose, above all the novels of Bāṇa (7th century) and Subandhu, and much technical and philosophical writing, uses compounds of great length. They must be analysed layer by layer. An illustrative example: nīlotpaladalaśyāmā is a bahuvrīhi, “(a woman) dark like the petal of a blue lotus”. Within it, nīlotpala (“blue lotus”) is a karmadhāraya; nīlotpaladala (“petal of a blue lotus”) a genitive tatpuruṣa; nīlotpaladalaśyāma (“dark as a blue lotus petal”) a compound of comparison.

6.4 The Case System

Sanskrit nouns have seven cases (vibhakti) in three numbers (singular, dual and plural). The vocative (sambodhana, “address”) is not a separate case in Pāṇini’s system but a use of the nominative (2.3.47–49); the special vocative singular form is called sambuddhi. Modern grammars usually list it as an eighth case (Tables 6.6, 6.7 and 6.8).

Table 6.6. The Sanskrit cases and their principal uses
No. Name Modern term Principal uses
1 prathamā nominative The bare meaning of the stem; the agent of an active verb, the object of a passive one
2 dvitīyā accusative The object (karman); goal of motion; duration
3 tṛtīyā instrumental Instrument (karaṇa); agent of a passive verb; accompaniment
4 caturthī dative Recipient (sampradāna); purpose
5 pañcamī ablative Point of departure (apādāna); cause; comparison
6 ṣaṣṭhī genitive “Other” relations, such as possession (not a kāraka)
7 saptamī locative Locus (adhikaraṇa) in space or time; the “locative absolute”
– sambodhana vocative Address; a use of the nominative
Table 6.7. Declension of deva “god” (masculine stem in -a)
Case Singular Dual Plural
Nominative devaḥ devau devāḥ
Accusative devam devau devān
Instrumental devena devābhyām devaiḥ
Dative devāya devābhyām devebhyaḥ
Ablative devāt devābhyām devebhyaḥ
Genitive devasya devayoḥ devānām
Locative deve devayoḥ deveṣu
Vocative deva devau devāḥ
Table 6.8. Declension of latā “creeper” (feminine stem in -ā)
Case Singular Dual Plural
Nominative latā late latāḥ
Accusative latām late latāḥ
Instrumental latayā latābhyām latābhiḥ
Dative latāyai latābhyām latābhyaḥ
Ablative latāyāḥ latābhyām latābhyaḥ
Genitive latāyāḥ latayoḥ latānām
Locative latāyām latayoḥ latāsu
Vocative late late latāḥ

Sanskrit has many more declension classes, distinguished by the final sound of the stem and by gender: stems in -i and -u, in -ī and -ū, in -ṛ (such as pitṛ “father” and mātṛ “mother”), and a variety of consonant stems, some of which change their form between “strong” and “weak” cases (rājā, rājānam, but rājñā, rājñaḥ). Pāṇini handles this variety with a single set of basic endings (sup) added to all stems, and a large number of rules that modify stems and endings in particular environments. Two further paradigms, a vowel stem in -i and a consonant stem in -an, are given in Table 6.9; the alternation of strong and weak stems in rājan is a feature that Sanskrit inherited from Indo-European. The kāraka system that underlies the assignment of all these cases is described in Chapter 5.

Table 6.9. Declension of agni “fire” (masculine stem in -i) and rājan “king” (masculine stem in -an)
agni rājan
Case Sg. Du. Pl. Sg. Du. Pl.
Nom. agniḥ agnī agnayaḥ rājā rājānau rājānaḥ
Acc. agnim agnī agnīn rājānam rājānau rājñaḥ
Ins. agninā agnibhyām agnibhiḥ rājñā rājabhyām rājabhiḥ
Dat. agnaye agnibhyām agnibhyaḥ rājñe rājabhyām rājabhyaḥ
Abl. agneḥ agnibhyām agnibhyaḥ rājñaḥ rājabhyām rājabhyaḥ
Gen. agneḥ agnyoḥ agnīnām rājñaḥ rājñoḥ rājñām
Loc. agnau agnyoḥ agniṣu rājñi, rājani rājñoḥ rājasu
Voc. agne agnī agnayaḥ rājan rājānau rājānaḥ

6.4.1 Pronouns

The pronouns have partly separate endings of their own. The demonstrative tad “that, he” has the dative, ablative and locative singular tasmai, tasmāt and tasmin, with an element -sm- that does not appear in the nouns, and the nominative plural te instead of -āḥ. (In Pali these pronominal endings spread to the nouns: Chapter 9.) The personal pronouns of the first and second persons have no gender, are built on different stems in the singular and plural, and have short unaccented forms (me, te, naḥ, vaḥ) that cannot begin a sentence (Table 6.10).

Table 6.10. Pronouns: aham “I”, tvam “you” and tad “that, he” (masculine), singular and plural
Case I we you (sg.) you (pl.) he they
Nom. aham vayam tvam yūyam saḥ te
Acc. mām asmān tvām yuṣmān tam tān
Ins. mayā asmābhiḥ tvayā yuṣmābhiḥ tena taiḥ
Dat. mahyam asmabhyam tubhyam yuṣmabhyam tasmai tebhyaḥ
Abl. mat asmat tvat yuṣmat tasmāt tebhyaḥ
Gen. mama asmākam tava yuṣmākam tasya teṣām
Loc. mayi asmāsu tvayi yuṣmāsu tasmin teṣu

6.5 Primary Derivatives (kṛt)

Suffixes added directly to roots form nouns, adjectives, participles, infinitives and absolutives. Pāṇini’s names for them include marker letters, which are shown in the second column (Table 6.11).

Table 6.11. Some primary (kṛt) suffixes
Suffix Pāṇini’s name Function Example
-ta kta past passive participle kṛta “done”, gata “gone”
-tavat ktavatu past active participle gatavat “having gone”
-at/-ant śatṛ present active participle pacat “cooking”
-māna/-āna śānac present middle participle pacamāna “cooking (for oneself)”
-tṛ tṛc agent noun kartṛ “doer”
-aka ṇvul agent noun pācaka “cook”
-ana lyuṭ action or instrument noun karaṇa “doing, instrument”
-tum tumun infinitive gantum “to go”
-tvā, -ya ktvā, lyap absolutive kṛtvā “having done”; prakṛtya (with a preverb)
-tavya, -anīya, -ya tavyat, anīyar, ṇyat gerundives kartavya, karaṇīya, kārya “to be done”

6.6 Secondary Derivatives (taddhita)

Secondary suffixes are added to nominal stems. Pāṇini treats them in Adhyāyas 4 and 5 (Table 6.12).

Table 6.12. Some secondary (taddhita) suffixes
Meaning Suffix Example
Patronymic -i (iñ), -a with vṛddhi (aṇ), -ya (yañ) dāśarathi “son of Daśaratha”; vāsudeva “son of Vasudeva”; gārgya “descendant of Garga”
Relation -ika (ṭhak etc.), -īya (cha) vaidika “Vedic”, dhārmika “righteous”, pāṇinīya “Pāṇinian”
Abstract nouns -tva, -tā manuṣyatva “humanity”, sundaratā “beauty”
Possession -mat/-vat, -in dhanavat “wealthy”, yogin
Material -maya suvarṇamaya “golden”
Time -ika dainika “daily”, māsika “monthly”
Place adverbs -tra, -tas tatra “there”, yatra “where”, tatas “from there”
Diminutive -ka putraka “little son”

Primary and secondary derivation, together with compounding, make the Sanskrit vocabulary highly productive and transparent: from the root kṛ “do, make” come kartṛ “doer”, karaṇa “instrument”, karman “action”, kārya “what is to be done”, kṛti “a work”, and with preverbs prakṛti “nature, base”, vikṛti “modification”, saṃskāra “refinement”, ākṛti “form”. This productivity is why Sanskrit has been a source of technical vocabulary for many Indian languages, including the coinages of modern Hindi, Bengali, Marathi, Telugu and Kannada.

6.7 The Verb

6.7.1 Tenses and moods (lakāra)

Pāṇini designates the tenses and moods by ten abstract elements, all beginning with l (hence lakāra), which are later replaced by the personal endings. The optative liṅ has two forms, the ordinary optative and the benedictive (Tables 6.13 and 6.14).

Table 6.13. The ten lakāras, illustrated with the root bhū “be, become” (3rd person singular active)
Lakāra Modern term Pāṇini’s characterisation Form
laṭ present present time (3.2.123) bhavati
liṭ perfect past, not witnessed by the speaker (parokṣa) babhūva
luṭ periphrastic future future, not of today bhavitā
lṛṭ simple future future bhaviṣyati
leṭ subjunctive Vedic only bhavāti
loṭ imperative command, request bhavatu
laṅ imperfect past, not of today abhavat
liṅ (vidhi) optative injunction, possibility bhavet
liṅ (āśīḥ) benedictive blessing bhūyāt
luṅ aorist past in general abhūt
lṛṅ conditional unrealised condition abhaviṣyat
Table 6.14. Present tense (laṭ) of bhū (active) and edh “prosper” (middle)
Active (parasmaipada) Middle (ātmanepada)
Person Sg. Du. Pl. Sg. Du. Pl.
3rd bhavati bhavataḥ bhavanti edhate edhete edhante
2nd bhavasi bhavathaḥ bhavatha edhase edhethe edhadhve
1st bhavāmi bhavāvaḥ bhavāmaḥ edhe edhāvahe edhāmahe

The Indian grammarians list the persons in the order third, second, first (prathama, madhyama, uttama puruṣa), as in the table; European grammars reverse the order. Table 6.15 gives the other common active forms of the same root.

Table 6.15. Other tenses and moods of bhū (active)
Imperfect (laṅ) Optative (liṅ) Imperative (loṭ) Perfect (liṭ) Future (lṛṭ)
3rd sg. abhavat bhavet bhavatu babhūva bhaviṣyati
3rd du. abhavatām bhavetām bhavatām babhūvatuḥ bhaviṣyataḥ
3rd pl. abhavan bhaveyuḥ bhavantu babhūvuḥ bhaviṣyanti
2nd sg. abhavaḥ bhaveḥ bhava babhūvitha bhaviṣyasi
2nd du. abhavatam bhavetam bhavatam babhūvathuḥ bhaviṣyathaḥ
2nd pl. abhavata bhaveta bhavata babhūva bhaviṣyatha
1st sg. abhavam bhaveyam bhavāni babhūva bhaviṣyāmi
1st du. abhavāva bhaveva bhavāva babhūviva bhaviṣyāvaḥ
1st pl. abhavāma bhavema bhavāma babhūvima bhaviṣyāmaḥ

Table 6.15 shows several characteristic devices of the Sanskrit verb: the augment a- that marks the imperfect (and the aorist and conditional) as past; the optative element -e- (historically -ai-, from -a- + -ī-); the reduplication of the root in the perfect (ba-bhūva); and the future suffix -sya-, here -iṣya-.

In classical Sanskrit the distinctions between the three past tenses were largely lost, and the past passive participle increasingly took over their functions, a development continued in the modern Indo-Aryan languages.

6.7.2 Active and middle (parasmaipada, ātmanepada)

Verbs take one of two sets of endings: parasmaipada (“word for another”) and ātmanepada (“word for oneself”). Some roots take only one set, others both (ubhayapada). For roots that take both, the traditional rule is that the middle endings are used when the fruit of the action accrues to the agent (yajate “he sacrifices for himself”) and the active endings otherwise (yajati “he sacrifices (for another)”), though in classical usage the distinction is often not observed. The passive is formed with the middle endings and the suffix -ya- (kriyate “is done”), introduced by rule 3.1.67 (sārvadhātuke yak).

6.7.3 Derivative verbs

Besides the simple root, Sanskrit forms several kinds of derivative verb stems, which Pāṇini introduces at the beginning of Adhyāya 3 and then treats as roots in their own right (3.1.32 sanādyantā dhātavaḥ):

Causative
with the suffix ṇic (3.1.26), for an action that the subject causes another to perform: bhāvayati “causes to be”, kārayati “causes to do, has made”, pācayati “has (someone) cook”.
Desiderative
with the suffix san and reduplication (3.1.7), for an action that the subject wishes to perform: bubhūṣati “wishes to be”, jigamiṣati “wishes to go”, pipāsati “wishes to drink, is thirsty”.
Intensive
with the suffix yaṅ and a strengthened reduplication (3.1.22), for an action repeated or done intensely: bobhūyate, jaṅgamyate “goes to and fro”.
Denominative
formed from nouns, for example with kyac (3.1.8) in the sense of wishing for something for oneself: putrīyati “wishes for a son”.

The causative is by far the most common of these in the classical language, and the desiderative supplies many ordinary nouns and adjectives (pipāsā “thirst”, jijñāsā “desire to know, inquiry”, mumukṣu “one who desires liberation”). The last two words show how the grammatical categories entered the vocabulary of philosophy: the Brahmasūtra opens with athāto brahmajijñāsā, “now, therefore, the desire to know brahman”.

6.7.4 Vowel gradation

Many Sanskrit forms alternate between three grades of a vowel: the basic grade, guṇa and vṛddhi (Table 6.16).

Table 6.16. Vowel gradation: basic, guṇa and vṛddhi grades
Basic grade guṇa vṛddhi
i, ī e ai
u, ū o au
ṛ, ṝ ar ār
ḷ al āl
a a ā

Pāṇini’s first two rules define vṛddhi and guṇa (1.1.1–2). The alternations they describe are of Indo-European origin, and the terms guṇa and vṛddhi were adopted by 19th-century European comparative philologists in describing the vowel gradation (ablaut) of other Indo-European languages (Chapter 14).

6.8 Vedic and Classical Sanskrit

Pāṇini described both the Vedic language (chandas) and the spoken Sanskrit (bhāṣā) of his day, and the differences between them are considerable. The Vedic language has a richer verbal system, including a subjunctive (leṭ) and an injunctive, and a dozen or more infinitive forms (such as -tave, -dhyai, -tos), where Classical Sanskrit has only -tum. Vedic nouns show more varied endings (for example the nominative plural devāsaḥ beside devāḥ, and the instrumental plural devebhiḥ beside devaiḥ). Preverbs may stand apart from their verbs (tmesis). Above all, Vedic has a pitch accent, which Pāṇini describes in detail and which was lost in Classical Sanskrit.

The Classical language that Pāṇini’s grammar fixed was used for over two thousand years with remarkably little change in its morphology. Its vocabulary, style and syntax, however, did change: later prose favours long compounds and nominal constructions, the past passive participle displaces the finite past tenses, and many words acquire new meanings or are borrowed from the Prakrits, the Dravidian languages and Persian. The grammar’s rules ensured the stability of forms; usage continued to evolve within them.

6.9 A Sentence Analysed

The pieces described in this chapter come together in the simplest sentence. Consider:

devadattaḥ kuṭhāreṇa vṛkṣaṃ chinatti
“Devadatta cuts the tree with an axe.”

  • chinatti is the third person singular active present of the root chid “cut”, a root of the seventh class, which inserts a nasal into the root to form its present stem (chi-na-t-ti). The active ending expresses the agent (kartṛ).

  • Because the agent is already expressed by the verb ending, the word for the agent, devadatta, takes only the nominative, which expresses “the bare meaning of the stem” (2.3.46).

  • vṛkṣam “tree” is the object (karman), what the agent most wishes to reach by the action (1.4.49), and takes the accusative (2.3.2).

  • kuṭhāreṇa “with an axe” is the instrument (karaṇa), the most effective means (1.4.42), and takes the instrumental (2.3.18); its ṇ results from the retroflexion rule after r.

  • Final -m before a consonant becomes anusvāra: vṛkṣaṃ chinatti.

In the passive the same participants are presented differently: devadattena kuṭhāreṇa vṛkṣaś chidyate “the tree is cut by Devadatta with an axe”. Now the verb (chid + passive -ya- + middle ending) expresses the object, so vṛkṣa stands in the nominative and the agent devadatta in the instrumental; the visarga of vṛkṣaḥ becomes ś before ch. Because case endings, not position, mark the relations, the words of either sentence may be rearranged for emphasis or metre without changing who does what to whom. This is the sense in which Sanskrit word order is “free”, though prose has a normal order (agent, object, verb) and departures from it are marked.

6.10 The Architecture of the System

Figure 6.2 brings together the stages by which the Pāṇinian system builds words and sentences.

A simplified picture of word and sentence formation in the Pāṇinian system
Figure 6.2. A simplified picture of word and sentence formation in the Pāṇinian system

The kāraka analysis used in this chapter has also been applied directly in the computational analysis of Sanskrit and Hindi, notably in the “Paninian Grammar Framework” of Akshar Bharati, Vineet Chaitanya and Rajeev Sangal (Bharati et al. 1995); its modern comparisons and limits are discussed in Section 5.7 and Chapter 14.

Further Reading

  • Cardona, George. 1997. Pāṇini: His Work and Its Traditions. 2nd ed. 1: Background and Introduction. Motilal Banarsidass.. — On the Pāṇinian analysis of Sanskrit.

  • Katre, Sumitra M. 1987. Aṣṭādhyāyī of Pāṇini. University of Texas Press.. — For the rules cited in this chapter.

  • Bharati, Akshar, Vineet Chaitanya, and Rajeev Sangal. 1995. Natural Language Processing: A Paninian Perspective. Prentice-Hall of India.. — On kāraka-based analysis in computational linguistics.