1. Introduction

Although its original reference is to the diachronic process by which two previously unrelated words or word stems come to be associated, resulting in a new, mixed paradigm (Boyé 2006), the term suppletion now more generally refers, in modern Western linguistics, to a synchronic alternation, a type of allomorphy in which the relation between allomorphs is not synchronically derivable. The English alternation between go and went is a well-known example.

In talking about suppletion, the focus is generally on stem suppletion: two or more synchronically unrelated stems form the basis of different parts of an inflectional paradigm, while suffixation is (usually) not directly affected. In principle, it would be possible to consider affix allomorphy (whether inflectional or derivational) in terms of suppletion. But since affix allomorphy rarely involves synchronically relatable affixes, this would almost completely subsume the notion of affix allomorphy under suppletion. In some cases, it may be necessary to treat full words as suppletive, where it is not possible, or would be unduly complicated, to decompose a form into suppletive stem and regular affix; Boyé (2006) calls this “inflectional form suppletion”. Suppletion is usually an inflectional phenomenon, but can feed into derivational formations, as in Latin lator based on the suppletive stem lat- of the verb ferre ‘bear’.

There is of course a gradient between ‘regular’ morphological relations and unambiguously suppletive relations. At the one end we have allomorphs which are synchronically relatable by means of productive phonological rules (such as noun stems ending in voiced vs. unvoiced stop in German: Tag [tak] vs. Tage [tag-]), at the other end allomorphs which share no relatable phonological content (like English go vs. went). There are a variety of different approaches for where to draw the line in terms of which alternations should be derived by phonological rules, and which treated as suppletion.

For a summary of issues in suppletion see Boyé (2006). For more detailed treatments see e.g. Dressler (1985), Mel’čuk (1994), Veselinova (2006), Corbett (2007).

2. The Indian tradition

The ancient Indian tradition does not have a term or concept precisely equivalent to suppletion, but its approach to suppletive and suppletive-like phenomena reveals a number of interesting comparisons with Western thought. First and foremost, all phenomena which would be considered instances of suppletion are treated, in the Indian tradition, in terms of substitution (ādeśa).[1] Substitution is one of the most important processes in Indian generative analysis, the means by which all allomorphy, and much allophony, is accounted for (including alternations of the German Tag–Tage type: Pāṇini would simply substitute [k] for [g] in the relevant forms).

Pāṇini’s Aṣṭādhyāyī provides no discussion of any conceptual difference between suppletion and other types of stem modification. There is, though, a practical distinction between stem modification (usually involving some kind of phonological substitution) which preserves some part of an original root, and stem modification in which an entire stem is replaced. For example, in the derivation of the present stems of the root kṛ ‘do, make’, namely karo- and kuru-/kurv-, the largely predictable morphophonological alternations are dealt with by means of substitution rules which have general application beyond merely this verb and these stems. For example, the vowel gradation which produces the root form kar (‘guṇa grade’) from kṛ is a general process found across the language, constrained by morphological criteria and obeying regular phonological principles. It is specified by means of a substitution rule targeting the vowel of roots in the affected contexts (Aṣṭ. 7.3.84).[2] Ultimately, these stem forms can be derived by means of generalizable rules involving common processes of stem modification such as suffixation and vowel gradation, formalized in terms of substitution. But crucially, at no point is the root as a whole (in whatever form) subject to substitution: the k of kṛ is never replaced, and thus in some real sense the stems karo- and kuru-/kurv- are modified forms of kṛ.

In contrast, the suppletive stem vadha- to the root (and stem) han ‘strike, slay’ is specified as a substitute for the root in its entirety in the relevant context (A. 2.4.42–44). As there is no generalization possible regarding the form alternation between han and vadha-, full substitution is the only option.

However, full substitution of a root/stem is also employed in cases where there are obvious phonological similarities between stems. For example, Aṣṭ. 7.3.78–9 specifies a series of stem substitutions:

• pā > piba-
• ghrā > jighra-
• dhmā > dhama-
• sthā > tiṣṭha-
• mnā > mana-
• dā > yaccha-
• dṛś > paśya-
• ṛ > ṛccha-
• sṛ > dhāv-
• śad > śīya-
• sad > sīda-
• jñā > jā-
• jan > jā-

In some of these cases, we are dealing with fully suppletive stem alternants, i.e. stem alternants with no phonological similarity or relationship, as in dṛś–paśya.  In other cases, a morphophonological relation between the root and the substitute is evident; for example, ghrā–jighra and sthā–tiṣṭha involve reduplication (as do pā–piba and sad–sīda, though more opaquely). Yet no distinction is made between these and cases like dṛś–paśya. It would certainly have been possible to frame a rule which derived, for example, tiṣṭha- from sthā and jighra- from ghrā (and perhaps most easily, ṛccha- from ṛ, simply by suffixation). Pāṇini’s method here appears to be based on the principle of concision, rather than any conceptual notion of the degree of relation one stem bears to another. This is illustrated by the preceding rule, Aṣṭ. 7.3.77, which provides a contrast to 78-9: this rule takes the roots iṣ, gam and yam, and produces the stems iccha-, gaccha-, and yaccha- (instances of partial suppletion); it does this by substituting the final segments of these roots with ccha, rather than by substituting the roots in full. In this case the morphophonological relation is recognized, because it enables the same alternation to be specified for three roots in one concise statement, more concisely than if the three substitute stems were to be given in full in the grammar.[3]

A particularly striking case is found in Aṣṭ. 2.4.41: the root veÑ [ve-] ‘weave’ is optionally substituted in full by what is superficially its sandhi variant vay [vay-], that is by a form which ought to be derivable from the base form of the root by fully productive and predictable phonological rules. The reason for this substitution is to prevent in the relevant contexts a further more general substitution which would apply to the unsubstituted vowel of veÑ to produce a stem vā-. The somewhat unintuitive outcome is that resulting words which contain forms of the root most phonologically close to the original root form (e.g. the perfect tense uvāya) derive from the substituted root, while those which contain root forms phonologically more distant from the original (e.g. the alternative perfect tense, vavau) do not involve full root substitution and thus preserve unsubstituted the first segment of the root.

One important point of dichotomy in the full root substitutions in the Aṣṭādhyāyī is their location in the grammar. Certain substitutions, including han–vadha and veÑ–vay, occur early in the grammar (in book 2 of 8), and these substitutions can feed later suffixation processes. Other substitutions are specified later, such as those in 7.3.78–79 discussed above; these substitutions do not feed other morphological processes, but are purely formal stem variants. For example, substitutions specified in book 2 may feed derivational morphology, such as the substitution of the copula as ‘be’ with bhū, which licenses derivational forms to this suppletive stem such as bhavitavyam, bhavitṛ. In contrast, no derivation is possible from stems like tiṣṭha-.

As discussed, suffix alternations are also treated in terms of full or partial substitution. In some cases fully suppletive words or even paradigms are produced by concomitant root and suffix substitution. Aṣṭ. 3.4.84 specifies the optional replacement of the root brū ‘speak, say’ with the stem āh-, together with concomitant replacement of present tense verbal affixes with (formally) perfect tense affixes, to produce forms such as 3sg. āh-a, 3pl. āh-uḥ ‘says’ as alternatives for brav-īti, bruv-anti ‘says’. In rare cases, whole words are substituted in their entirety, where no morphological segmentation of the resulting form is either possible or profitable. This is the case with the enclitic forms of the personal pronouns (Aṣṭ. 8.1.20-23), e.g. 2du. acc./dat./gen. vām replaces the whole of the basic forms yuvām/yuvā-bhyām/yuva-yoḥ.

A further important point is the centrality of meaning to Pāṇini’s procedure, which stands alongside concision as one of the core driving features of his analysis. Returning to āha: is this really an instance of suppletion, given that it is optional? Or is optional suppletion merely synonymy? In terms of a paradigmatic morphology, there is no slot in the paradigm of brū which āha fills, or at least, its slot is already filled by bravīti. For Pāṇini the rationale is driven by meaning: āha is synonymous with bravīti, and it is only on this basis that one can be treated as a substitute for the other.  Similarly, the comparative śreyas ‘better’ and superlative śreṣṭha ‘best’ are not derived, as historically appropriate and phonologically possible, from the word śrī ‘radiant, holy’, but from the historically unrelated, and phonologically more distant, adjective praśasya ‘good, praiseworthy’ (Aṣṭ 5.3.60). The rationale is entirely semantic: Pāṇini cannot treat śreyas and śreṣṭha as comparative and superlative of śrī, despite their evident connection, because they no longer function semantically as regular comparative and superlative of this word. Rather, their meanings are effectively comparative and superlative of the meaning of praśasya, so Pāṇini treats them as derived from this adjective.

Further on the linguistic reality of verbal suppletion in Sanskrit, and its relation to the full verb stem substitutions specified by Pāṇini, see Deshpande (1992).

[1] The use of the term ādeśa to mean ‘substitute, substitution’ is a feature of the Pāṇinian grammatical tradition, but this was not, as argued by Acharya (2017), Pāṇini’s own use of the term.

[2] By Aṣṭ. 7.3.84 (together with 1.1.3), the vowel a replaces the ṛ of kṛ, giving an intermediate form ka, to which is necessarily appended an r, by Aṣṭ. 1.1.51.

[3] Aṣṭ. 7.3.77 is seven syllables long; if this rule were eliminated and the three substitutions were incorporated into the following rule, it would require 7.3.78 to be at least eleven syllables longer.