A compound can be defined in very basic terms as a word that is made up of two or more words. Compounds thus show properties of both morphological formation (the whole constitutes a word) and syntactic formation (the whole is constituted by two or more related words), to the extent that there is an ongoing debate on which component of the grammar is responsible for their formation (Bauer 2006).

Compounding is a highly productive process in Sanskrit. The Indian grammatical tradition focused primarily on the process of compound formation, particularly with a view to how transparently, or otherwise, the meaning of the compound can be derived from the meaning of its parts. The better known subject of Indian grammatical thought on compounding, however, at least within Western linguistics, is that of the classification of compounds, an area where Indian influence on the West is at least superficially obvious, in the introduction of classificatory terms such as bahuvrīhi, dvandva, etc.

1. Definition and properties of compounds

Samāsa is the technical name employed in the Indian grammatical tradition to designate both the process of compounding and the form obtained by such a process. Compounding as an operation applies to fully inflected words which are semantically connected, not just juxtaposed, according to Aṣṭ. 2.1.1 samarthaḥ padavidhiḥ, “an operation (vidhi) concerning fully inflected words (pada) applies to words which are semantically connected (samartha)”.[1] Pāṇini devotes a large section to compounding in Aṣṭ. 2.1.1-2.2.38.

Compounding two or more words (pada) results in a prātipadika 'nominal stem', which is also the output of the major derivational morphological processes in Sanskrit, such as kṛt and taddhita suffixation. Being a nominal stem, the compound as a whole receives a case-affix and thereby itself attains the status of word (pada); see example (1).

The elements entering compound formation are fully inflected words complete with case-endings (or person/number endings, in the rarer case of compounded verbs). The case-endings of the constituent words are standardly deleted in the process of compound formation, as shown in (1). In certain irregular compounds, known as aluksamāsa ('non-deletion compounds'), the first member may retain its original case-ending, as shown in (2).

As a nominal stem, any compound can be subject to further derivation. For example, the suffix -vat is shown suffixed to an underived word in (3a), but it can also be assigned to a compound stem, such as asattva-vacana- in (3b). Any compound can also be further compounded with another word, as shown in (4); this potential for recursion is productively employed, leading to very long compounds.

Sanskrit compounds display syntactic properties that are not commonly found with compounds in other languages. For instance, Patañjali (on Aṣṭ 2.2.1, Mbh. 1.360.20) notes that Sanskrit allows constructions called asamartha or 'non-constituent' compounds, in which there is a syntactic relation between a subordinate compound member and a word outside the compound. This possibility is generally excluded in other languages (Di Sciullo and Williams 1987).[2] For example, in (5) the compound-external word devadattasya is construed with the word guru, which is a subordinate member of the compound; it is not construed with the compound as a whole.

Another interesting syntactic property of Sanskrit compounds is their potential lack of anaphoric islandhood. All Sanskrit pronouns have forms for use in compounds, including the demonstrative and relative pronouns, which usually refer to elements outside the compound in which they appear, e.g. (6) (Lowe 2015).

This particular property has not received special attention in the Indian grammatical tradition, since it follows the regular pattern of compound formation in Sanskrit, but it goes beyond the possibilities available in most other languages, including English. The fact that subordinate elements of Sanskrit compounds can entertain syntactic relations outside the compound is evidence for the syntactic, rather than morphological, formation of compounds in Sanskrit, according to Lowe (2015). Nevertheless, within the Indian tradition, the result of compound formation is unquestionably the same as the result of other morphological processes, namely a word (pada).

2. Classification

There is a large variety of compound types in Sanskrit, though many are relatively rare. Pāṇini classified compounds into four broad classes, which are described in detail in Aṣṭ. 2.1 and 2.2. These classes are labelled: tatpuruṣa, bahuvrīhi, dvandva and avyayībhāva.

The term tatpuruṣa covers a wide range of compounds, but the canonical type involves the implication of a case relation between the first element (a noun) and the second element (noun or adjective), as in (1), (2), (4) and (7). This type of compound is termed “determinative” by Whitney (1889); modern Western classifications also use the term “subordinative” (or “subordinate” for Scalise and Bisetto 2009).

An important subtype of tatpuruṣa is karmadhāraya, the most common varieties of which are the combination of an adjective with the noun it modifies, e.g. (8a), the combination of adjective and adjective/adverb or adverb and adjective, where the first member modifies the second, e.g. (8b), and the combination of two nouns which both refer to the same entity(es), e.g. (8c). Karmadhāraya compounds are called “descriptive” by Whitney; in the terms of Scalise and Bisetto (2009) they mostly fall under the type “attributive/appositive” by Scalise and Bisetto, but also (in the type in 8c) under the type “coordinate”.

Bahuvrīhi compounds can be thought of as a kind of reduced relative clause. A common type is an adjective followed by a noun, where the referent of the compound is an entity which is in some way described by the adjective+noun sequence. So the compound in (9) refers to some entity whose ears are long. Bahuvrīhis are called “possessive” compounds by Whitney, but possession is only one of the possible relations between the head of the compound and the referent of the compound. They have largely been equated with the term “exocentric” in Western linguistics, but again exocentric is a broader category than the Sanskrit category of bahuvrīhi (Bauer 2017).[3]

Dvandva compounds are compounds of coordinate nouns. There are two types: itaretara dvandva, e.g. (10a), and samāhāra dvandva, e.g. (10b). In the first type, the compound adopts the gender and declension of the final member and its number, dual or plural, depends on the sum of the number of its constituent members. In the second type, the compound is invariably singular and neuter; the elements referred to by the compound are seen as a composite unit.[4] Dvandva compounds are called “copulative” by Whitney (1889); “coordinative” or “coordinate”[5] are largely corresponding terms in the Western tradition.

Finally, avyayībhāvas are compounds involving a preposition and a governed noun, functionally equivalent to an adverbially used prepositional phrase, e.g. (11). This class is not much discussed in Western classifications, probably because other languages do not have a distinctive class defined by precisely the features that define it in Sanskrit (Bauer 2017).

Patānjali proposes that Pāṇini's fourfold classification of compounds can be justified semantically, based on the predominance, pradhānatva, of the members of the compound. So, in avyayībhāva compounds the first member is semantically predominant; in tatpuruṣas the second member is predominant; in dvandvas both members are predominant, and in bahuvrīhis an external element is predominant. This notion of predominance thus effectively represents a notion of semantic headedness, and on some level relates to the distinction between endocentric and exocentric compounds. This cross-classification can be interpreted as a four-way classification based on the interaction of two binary features:

However, the later grammarians Bhaṭṭojidīkṣita and Kauṇḍabhaṭṭa show that this semantic interpretation of Pāṇini's categories is flawed; all four categories include compounds which do not fit the semantic definition proposed by Patañjali. Kauṇḍabhaṭṭa's conclusion is that these categories have no semantic definition, but are mere arbitrary labels applied by Pāṇini for the convenience of his grammatical system.

The modern Western tradition is familiar with only the four-way division of compounds given by Pāṇini, but in fact there are a number of competing and complementary classificatory systems in the Indian tradition. For example, Bhaṭṭojidīkṣita's Vaiyākaraṇasiddhāntakārikā 28 presents a purely morphological classification based on the types of words that combine in a compound. According to Aṣṭ 2.1.4 saha supā, a word that ends in a nominal ending (sUP) compounds with another word ending in a nominal ending (sUP). Bhaṭṭojidīkṣita, however, refers to six possible combinations: sUP 'inflected noun' + sUP 'inflected noun', e.g. (12a); sUP 'inflected noun'+ tiṄ 'inflected verb', e.g. (12b); sUP 'inflected noun' + nāman 'nominal stem', e.g. (12c); sUP 'inflected noun' + dhātu 'verbal root', e.g. (12d); tiṄ 'inflected verb' + tiṄ 'inflected verb', e.g. (12e); and tiṄ 'inflected verb' + sUP 'inflected noun', e.g. (12f).

A different, two-fold, morphological classification is also found: luksamāsa vs. aluksamāsa. The first type refers to regular compound formation in which the case-affix appears only at the end of the compound, e.g. (1) or (12a); the second type refers to those irregular compounds that retain the case affix in the first member, as shown in (2) above.

A two-way classification based on semantics, rather than morphology, is given by Jayāditya (Murti 1974: 78). According to the Naiyāyika Jagadīśa, Jayāditya classified compounds primarily as either nitya (obligatory) or anitya (optional). According to Patañjali, the meaning of a nityasamāsa cannot be expressed with an analytical phrase, while the meaning of an anityasamāsa corresponds directly to the meaning of the analytical phrase from which it is derived. Take, for example, the compound in (13a). Its sense can be expressed by the phrase in (13b), making it an anitya compound.

However, the sense of the compound in (14a) cannot be expressed by the phrase in (14b) because its meaning is more specific than that of the analytic phrase, referring to a particular species of snake. The same is true for the compound in (15a), which refers to a particular species of rice, without any necessary attribution of the colour red.

These examples involve semantic specialization, but other factors may serve to make a compound nitya, for example the upapada compounds where the second member cannot be used as an independent word.

These and other Indian classification systems for compounds are discussed in detail by Murti (1974).

3. Compound meaning: ekārthībhāva and vyapekṣā

While the (or rather, an) Indian approach to compound classification is well known in the Western linguistic tradition, perhaps the more central object of thought in the Indian tradition was the process of compound formation, in particular the semantic process and the relation between the meaning of the compound and that of its constituent members.

The view of Kaiyaṭa interpreting Patañjali is that there are four possible approaches to meaning in compounding (cf. Murti 1974):

Table 1: Four possible views of compounding

In the first view (nityapakṣa / nityaśabdavāda), compounds are independent units with underived meanings; their derivation from phrases is a grammatical fiction. This reflects a broader philosophical stance within the grammatical tradition which held that words and even whole sentences or utterances are, in reality, unitary, and that grammatical derivation is unreal. This view is most strongly advocated by the grammarian-philosopher Bhartṛhari, but at least in relation to compounds this view significantly antedates him, being evident in the work of Patañjali, for whom it may have been the preferred approach (so Murti 1974).

In contrast, under the vṛttipakṣa / kāryaśabdavāda, compounds are in some real sense modifications of syntactic units, and the compound meaning is derived in some way from the meanings of the words that constitute it. The simplest of the views falling under this heading is that of vyapekṣā: the meaning of a compound is essentially the same as the meaning of the syntactic phrase from which it derives. The combination in compound denotes the relation (which in the phrase is denoted by the explicit syntactic relation, marked by case endings, etc.).

Under ekārthībhāva, compounds are morphologically and semantically derived, as under vyapekṣā, but here the meaning of compounds is understood to be different, and specifically unitary, in comparison with the meaning of the corresponding analytical phrase. Within this there are two possible approaches to the semantic process by which the unitary compound meaning is derived: i) jahatsvārthā vṛtti and ii) ajahatsvārthā vṛtti. The process (vṛtti) can be understood to be jahatsvārthā '(wherein) the words' own meanings are abandoned'; that is, in a compound like rājapuruṣa 'king's servant' (12a), the constituent words rājan and puruṣa do not continue to express their respective meanings in the compound, but denote only the unitary (and therefore separate) compound meaning.

Alternatively, the process can be understood as ajahatsvārthā '(wherein) the words' own meanings are not abandoned'; on this understanding the words in a compound continue to express their own meanings, while also denoting the unitary (and therefore separate) compound meaning.

This is far from the only approach to classifying possible approaches to the compounding process in the Indian tradition. A rather different approach is advanced by Kauṇḍabhaṭṭa, who presents an innovative interpretation of Patañjali in which jahatsvārthā vṛtti is equated with ekārthībhāva, and ajahatsvārthā vṛtti with vyapekṣā. Kauṇḍabhaṭṭa rejects the nityapakṣa as a viable option, and thus reduces the four-way inventory of possible viewpoints to two.

4. Ekaśeṣa

See the Coordination and Ellipsis entries.

[1] The precise meaning of this sūtra has been discussed in the tradition. For instance, samartha is interpreted as “capable”, i.e. a compound is capable of reflecting the meaning of the syntactic phrase.

[2] These constructions have been examined within different frameworks in the literature (Lowe 2015, Molina-Muñoz 2013, Gillon 1994, Kiparsky 1982, among others).

[3] See Pontillo (2021) for an in-depth discussion of exocentric compounds in Western linguistics and the re-interpretation of Aṣṭ. 2.2.24 anekam anyapadārthe.

[4] For a more detailed description on the multiple types of coordination, see the “Coordination” entry in the LINGUINDIC database.

[5] Whitney (1896: 488) refers to āmreḍita constructions under copulative compounds, headed “Repeated Words”, and several modern scholars classify them as iterative compounds (Delbrück 1893, Renou 1952). In fact āmreḍitas are not considered compounds by Pāṇini but merely iterated words. However, Śākalya in the Ṛgvedapadapāṭha treats them as compounds (Ditrich 2011).