Research on relative clauses in modern Western linguistics includes both more descriptive/typological angles such as distinguishing the different semantic and syntactic types of relative clause attested cross-linguistically, and more theoretical angles dealing with the formal syntactic and semantic analysis of relative clauses and their relation to the elements they modify and the matrix clauses they appear within. For a summary, see Nikolaeva (2006); for more detailed discussions see Alexiadou et al. (2000), de Vries (2002).
Work on relative clauses in modern Western linguistics has been considerably influenced by the predominance of a particular relativization strategy in languages of the European linguistic area, in particular English. This strategy involves an embedded clause, which is syntactically subordinated to the head it modifies, and which contains, usually in first position, a relative pronoun coindexed with a gapped argument position in the embedded clause:
(1) The man who you thought I knew
In (1), the relative clause ‘who you thought I knew’ is embedded within the noun phrase and subordinated to the head of the noun phrase, ‘man’. The relative pronoun ‘who’ begins the relative clause, and is coindexed with the gapped object position of ‘knew’.
As discussed by Comrie (1998), this relativization strategy is quite rare outside the European linguistic area, yet it has had a considerable influence on the development of analyses of relativization in modern Western linguistics, and on the analysis of other, typologically more common, relativization strategies.
2. Relativization in Sanskrit
In Sanskrit and other Indo-Aryan languages, the major relativization strategy is the relative-correlative construction. Correlatives are typologically rare, often limited to head-final languages (Downing 1973, Keenan 1985, de Vries 2002, Belyaev & Haug 2020). In a relative-correlative construction, the relative clause contains a relative pronoun, and is adjoined at the left or right edge (in Sanskrit standardly the left) of the main clause, which usually contains a correlative pronoun; see (2).
Sanskrit relative clauses are not embedded, but conjoined or symmetrically adjoined (Hock 1989, Davison 2009). Extensive work on the properties of correlative constructions has been undertaken on Hindi (e.g. Srivastav 1991, Bhatt 2003, among others).
The differences in relativization patterns between Sanskrit and the European languages central to the development of modern linguistics are reflected in the rather different approach to analysing relative constructions found in the Indian tradition.
3. The Aṣṭādhyāyī and the mainstream tradition
The analysis of relative clauses is not a central issue within the mainstream Indian linguistic tradition. Pāṇini’s Aṣṭādhyāyī does not explicitly treat any syntactic rules specific to the formation and use of relative clauses. In Sanskrit, finite verbs in relative clauses and certain other subordinate contexts are necessarily accented, in contrast to main clause verbs which are deaccented when non-initial. Yet even this morphophonological aspect is specified entirely formally, with reference to cooccurrence of forms of the relative pronoun (yadvṛtta, Aṣṭ. 8.1.66), rather than with reference to any functional properties of the clause. Similar is the specification of the optative mood for the verb in a specific semantic subset of subordinate clauses (Aṣṭ. 3.3.148-151).
In defining the scope of the absolutives, Pāṇini does refer to some range of the function of relative clauses. The absolutives are specified for use in referring to an action which precedes the action of a following verb, and which has the same agent (kartṛ) (Aṣṭ. 3.4.21). But by Aṣṭ. 3.4.23 this is not the case when the prior action occurs in a clause with a form of the relative yad, unless there is another verb which effectively serves as the main predicate of the yad clause. In the absence of the absolutive suffixes, the verb form in the yad clause will be finite, and we will therefore have a relative clause. In other words, Pāṇini’s specification for the absolutives shows that the scope of absolutives and relative clauses overlap in the case of reference to a prior event, e.g. ‘having done this, he returned’ ≈ ‘when he had done this, he returned’.
While nothing more specific is said on finite relative clauses, Pāṇini does specify syntactic/semantic restrictions on the formation of present participial clauses (illustrated in 3), which can be thought of as a type of reduced relative clause.
Aṣṭ 3.2.124-126 specify the restrictions holding between the head of the relative clause and the participle, namely that the referent (≈subject) of the participle is necessarily coreferent (samānādhikaraṇa) with an element in the sentence which is either non-nominative, in which case the participle may denote an event cotemporaneous with that of the main clause (as in 3), or, if the coreferent element is nominative, then there must be an additional semantic implication (such as cause) in the relation between the event denoted by the participle and that denoted by the main verb. For further details, see Lowe (2015: 329-335).
More broadly, the Indian tradition did not draw the same distinction between main and subordinate clause as found in the Western tradition. Perhaps the most closely related notion is the contrast between pradhāna-kriyā ‘predominant/primary verb/action’ and guṇa-kriyā ‘qualifying verb/action’, which might appear to represent a step towards distinguishing main from subordinate clauses, but in fact serves only to distinguish finite from non-finite clauses. Finite verbs are always pradhāna-kriyā; guṇa-kriyā is used only in relation to non-finite verb forms like absolutives and participles, whose subject positions are dependent on an argument from their matrix clause. The concept of guṇa-kriyā is therefore closer to modern notions of control in the broadest sense.
4. The Samanvaya tradition
In the Samanvayadiś and Samanvayapradīpa we find a more syntactically oriented discussion of relativization. Section 5 of the Samanvayadiś treats relations between clauses, and the treatment of relative clauses within this appears to be based on the assumption that there is no necessary hierarchical relationship between what in Western thought would be considered the main clause and the subordinate clause. This reflects the adjoined nature of relative clauses in Sanskrit. Rather, both clauses are self-contained, and in that sense independent, but they are connected due to a particular semantic relation existing between an element in each clause.
Building on this assumption, the most basic or most complete means of expressing such a relation between clauses is when both clauses contain an explicit marker specifying the relation. The marker in one clause is the relative pronoun; the marker in the other is a corresponding demonstrative (correlative) pronoun. Thus, the relative-correlative structure (as in ex. 2) is taken to be the paradigm type of interclausal relations.
In the relative-correlative case, the relation between the clauses is described as śābda, ‘determined by the words’, that is it is explicitly marked (Samanvayadiś 126.96.36.199 [Slaje’s numbering]). When at most one of the words expressing the connection between the clauses is present, we are dealing with relations which are ārtha ‘determined by the meaning’, that is contextually inferable but not explicit (SD 188.8.131.52).
Logically, there are two possibilities: either the demonstrative pronoun or the relative pronoun is missing. In the former case, that is when a relative pronoun occurs in one clause, but there is no explicit demonstrative/correlative in the other, we are dealing with something more akin to the standard English-type relative clause. The Samanvayadiś distinguishes two subtypes of relative-only clausal connections. One is labelled kalpita-karmādi-viṣaya ‘having as its range an inferred (demonstrative functioning as) patient or other grammatical role’ (SD 184.108.40.206.2), i.e. the relative is interpreted as being connected with a demonstrative pronoun which is inferred in a particular grammatical function. That is, in principle relative pronouns are understood as indicating a connection with another clause, a connection which can standardly only be picked up by a demonstrative pronoun in the other clause; in the absence of such a demonstrative, we infer an unexpressed demonstrative to pick up the connection with the relative. This first subtype refers to a relative-correlative structure in which the correlative is ellipsed or implied. For the following example, the Samanvayapradīpasaṃketa infers a correlative sa preceding rākṣasendraḥ in the main clause:
In the second subtype of clausal connection which involves only a relative pronoun, the relative is called prakrānta-vastu-viṣaya ‘having as its range an object previously mentioned’ (SD 220.127.116.11.2). This is the familiar type of non-restrictive relative clause, where a relative clause functions as modifier to a noun whose reference is already fully specified in the context.
Alongside clausal connections which involve only the presence of a relative pronoun, the Samanvaya treatises set as parallel those clausal connections which involve only the presence of a demonstrative pronoun (SD 18.104.22.168.1). All uses of the demonstrative/anaphoric pronoun tad are categorized under this heading, including its use to refer to something outside the immediate linguistic context but in the shared world knowledge of the interlocutors. This takes us beyond the scope of relative clauses from the Western perspective into the scope of anaphora and pronominal reference, but from the perspective of the Samanvaya tradition, there is no fundamental difference between the two: both are means of indicating a semantic relation between two otherwise independent clauses, and both relate in parallel ways to the fully explicit means of indicating relations between two independent clauses, i.e. the relative-correlative construction.
The typology of interclausal relations as understood in the Samanvaya tradition can be schematized in the following way:
The Samanvaya tradition also specifies a number of pragmatic constraints on acceptable usage of relative clauses, which go beyond the present scope. Similar and related observations on the use of relative clauses are found in Alaṃkāraśāstra, including in Mammaṭa’s Kāvyaprakāśa and the Vyaktiviveka.
 See Hahn (2008: 215-216). In Mammaṭa’s Kāvyaprakāśa, (VII, fllg. v. 187) it is claimed that the relative pronoun yad can be used without the demonstrative only when it occurs in the later of two connected clauses (i.e. from our perspective when following the main clause), not in the earlier.
 This schematization is slightly simplified; a more complex categorization is put forward by Ruyyaka, commentator on Mahimabhaṭṭa’s Vyaktiviveka, with additional mixed categories, as well as a category where neither relative nor demonstrative is explicit. See Hahn (2008: 58—71, esp. 68—69) for details.