In this study, I examine the historical development of focus concord constructions – that is, constructions similar to Japanese kakarimusubi, where the verb appears with a special affix and a focus particle usually appears somewhere within the clause – in Sinhala, an Indo-Aryan language of the Indo-European family spoken in Sri Lanka; with comparison to similar constructions in neighbouring South Dravidian languages Tamil and Malayalam.

The comparison of Sinhala with Tamil and Malayalam here is not simply due to a desire for crosslinguistic coverage nor simply because they are all three South Asian languages. Sinhala has been effectively cut off from other Indo-Aryan languages for roughly two millennia and has for the same period of time been in contact with South Dravidian languages, specifically Tamil. It has in fact been suggested that the morphosyntactic structure of focus constructions in Sinhala is essentially a calque of South Dravidian focus constructions (Gair 1986[1998]b). The examination of the Sinhala focus construction at various stages, compared to both early and modern South Dravidian focus constructions shows that – even if Sinhala did in essence borrow a Dravidian-like focus construction – the underlying properties of focus concord structures in Sinhala and South Dravidian differ not only in their earlier stages but in fact seem to show increased divergence by the modern period.

This paper is concerned with various aspects of the evolution of focus-concord constructions, including: (i) the source of focus-concord verbal forms; (ii) the source of at least some of the focus-related particles; and (iii) how the relationship between focus-concord verbal forms and focus-related particles changes over time, and what these relationships suggest with respect to the changes in the structure of focus-concord constructions diachronically.

In Sinhala focus-concord verbal forms originate from special-case uses of the “impersonal” nominalised construction in Old Sinhala, and develop from the originally “impersonal”/“existential” nominalised construction to a structure in Classical Sinhala where the nominalisation involved also creates an additional predicative structure, by putting the nominalised clause into a copular relation with the focussed element.1 By the stage of Modern Colloquial Sinhala these structures have been reanalysed as another sort of monoclausal structure. In the Dravidian languages examined (Tamil and Malayalam), on the other hand, focus-concord structures appear to involve predicative clefting (as in the Classical Sinhala structures) throughout extant texts.

Special attention is also paid to the focus concord particles (e.g. kakari-equivalents) which occur at different stages of Sinhala, Tamil, and Malayalam, for – at least in Sinhala – the focus construction (the musubi equivalent) can be clearly shown to have originated as independent of focussing, and not as immediately connected to the “focus” (kakari-type) particles. It is rather that the association of the predecessor of the later true focus construction in Sinhala was indeed not originally specific to focus, but in fact only became strongly associated with focus and with the use of focus particles, including question/quantifier particles [henceforth Q-particles],2 at a later stage. Examining the interaction of focus concord verbal forms and focus particles, particularly Q-particles, involves at some points the examination of occurrence of these particles in environments where they do not co-occur with a focus concord verbal form. Such cases are included in order to provide a fuller picture of the interplay of Q-particles and focus concord constructions.

This study therefore lays out a detailed examination of the history of the Sinhala focus concord construction, expanding upon the discussion of the evolution of Sinhala focus concord constructions in Slade (2011, 2013). I provide an overview of the evolution of focus concord constructions in Sinhala, and a somewhat less complete sketch of focus concord constructions in Dravidian – focussing on the South Dravidian languages Tamil and Malayalam – which can be usefully compared to the diachronic development of focus concord constructions in other languages, such as Japanese kakarimusubi.

Especially as concerns historical development and the linkage between focus concord constructions and focus particles, of particular interest in Sinhala and Dravidian is the original independence of kakari-type focus particles from the musubi-type focus concord construction (a situation which persists within present-day Dravidian), in contrast to the apparent situation in the earliest extant Japanese kakarimusubi constructions, where kakari particles originate as focus-associated particles which obligatorily trigger musubi verbal forms, with particles like ka only later developing into a true Q-particle.

The paper is organised as follows. Section 2 introduces the basic structure of the focus-concord construction in modern and Classical Sinhala, with comparison to Japanese kakarimusubi constructions. Section 3 examines the origins of the Sinhala focus-concord construction in Old Sinhala “impersonal” nominalised constructions. Section 4 pursues details of the development of focus-concord constructions in Sinhala. Section 5 examines the synchronic and diachronic distribution of (often focus-associated) Q-particles in Sinhala. Section 6 examines focus-concord constructions and focus-associating particles in Dravidian, focussing on Malayalam and Tamil (both modern and early). Section 7 compares the development of focus-concord constructions and Q-particles in Sinhala and Dravidian. Section 8 provides a general discussion of focus-concord constructions in Sinhala and Dravidian, with comparison to similar constructions in Japanese and other languages.

Modern Colloquial Sinhala [MCS] possesses a morphologically-overt focus construction in which the finite verb appears with an -E “focussing” ending rather than the usual -A “neutral” suffix. The focussed element in the clause optionally (but preferably) appears dislocated right of the verb. Also optional is the use of a focus particle like -y immediately following the focussed element. Compare (1a), which contains no focussed elements and thus uses the neutral form of the verb, with (1b), where potə ‘book’ is focussed and where the verb appears in the -E focus concord form. The example in (1b) shows the typical rightward dislocation of the focussed element. Example (1c) shows that even when the focussed constituent remains in situ, it still triggers the focus-concord verbal form (Slade 2011: 44–47).3

Modern Literary Sinhala [MLS]4 is similar to its colloquial counterpart, except that (a) dislocation is obligatory; (b) the copula/form of ‘to be’ is obligatory; (c) subjects of focussed verbs appear in accusative case; and (d) focussed verbs display no overt subject-verb agreement (in contrast to non-focussed verbs).5 Compare the non-focussed (2a) with its focussed counterpart in (2b).

As has been noted elsewhere, the focus-concord construction of Sinhala is reminiscent of the kakarimusubi construction found in early Japanese (see Sansom 1928; Ogawa 1976, 1977; Whitman 1997; Hagstrom 1998: 24–28; Watanabe 2002; Yanagida 2006; Aldridge 2009, amongst others). Both constructions involve a clause-internal (rather than clause-final) particle which induces a special marking on the verb. An example of an Early Middle Japanese kakarimusubi construction is given below in (3).

Similarly – as do focussing/emphatic particles such as koso, zo, and namu – we find that in Old Japanese the Q-particle ka also participates in kakarimusubi, as in (4).

So too, in Sinhala, Q-particles like də trigger the use of the focus concord -E verbal forms, as shown in (5).

However, as discussed by Serafim & Shinzato (2000); Watanabe (2002); Aldridge (2009); Mitrović (2014a), amongst others, Japanese ka seems to have originated not as a Q-particle required by interrogatives, but rather as a focus particle (which could appear in questions and was associated with interrogation) and only later became (largely) obligatory in interrogatives. Old Japanese questions therefore can occur without ka, as in:

On the other hand, the Sinhala particle də/da, as discussed in Section 5, does originate as a Q-particle, apparently originally associated specifically with alternative questions, but did not originally trigger focus concord. The roots of what becomes a focus concord construction is traceable back to the era of the Old Sinhala of the Sihigiri graffiti (8th–10th c. A.D.) – examined in more detailed in Section 3. Not only is this construction not originally triggered by the use of a Q-particle, the construction itself is not specifically associated with focus in Old Sinhala.

Focus constructions in Classical Sinhala [CS] (ca. 12–15 c. A.D.) closely resemble those of Modern Literary Sinhala [MLS], differing in that: (a) focussed elements can be dislocated either to the right-edge, following the verb, as in (7a), or to the left-edge of the clause, as in (7b); and (b) focussed subjects, unlike in MLS, continue to control verb agreement, as in (7c).7

Old Sinhala presents a rather different pattern: the properties of the focus concord construction in Old Sinhala, the earliest form of Sinhala with significant texts, dating from the 8–10 c. A.D., are different enough from that of the other stages as to warrant its own section, and this account forms the body of the following section.

In this section I show that the Sinhala focus concord originates in the Old Sinhala impersonal nominalisation construction which happens to be compatible with – but does not require – the presence of a focussed element. At this stage the focus concord construction does not have the predicative “clefting” character it does in its later Classical or Literary guises.

In the Old Sinhala [OS] of the graffiti on the Mirror Wall at Sihigiri (ca. 8–10 c. A.D.), there is not yet a specialised focus construction, but we can identify the roots of what would later become the focus concord construction. In OS, it is to be noted that there are various ways of nominalising verbs: one of these nominalisations is used to form both participant-nouns (e.g. ‘go-er’ from ‘go’, as in example (8)) and “impersonal” nominalised verbs (Slade 2011). The latter of these is of interest for us for it is the “impersonal” usage of the nominalised verbal form that sometimes co-occurs with focussed elements and thus appears to constitute the origin of later Sinhala focus concord constructions.

Salient properties of OS clauses containing “impersonal” nominalised verbs include: (a) “subjects” of impersonal nominalised verbs, when expressed, appear in the genitive case; (b) no overt agreement element or form of ‘to be’ is required. (However, when such an element does appear, it always immediately follows the nominalised verb, rather than following the focussed element as is the case in later Sinhala); (c) focussed elements, when present, are not obligatorily dislocated.

Example (9) provides an example of an OS clause utilising impersonal nominalisations where there is no apparent focussing of any element. In fact, both impersonals in (9) seem to involve backgrounding. That is, the impersonal nominalisation here seems to foreground the verbal elements (‘remaining’, ‘speaking’) by backgrounding the agents.

Nor at this time does ya/yi, the precursor of the modern colloquial Sinhala -y, yet behave as a focus particle, rather it functions as a sort of agreement clitic,9 standing in complementary distribution with the copula. This is unsurprising, given its etymological origin in the Old Indo-Aryan third-person singular present agreement morphology -ti (see Geiger 1941: 142; Karunatillake 2012: #9114).

Examples (10a)–(10d) illustrate ya/yi’s status as an agreement clitic rather than anything like a focus particle. In all of these examples, ya/yi occurs immediately following the nominalized verb, and in none of these examples does there appear to be any specially highlighted element.

In no case can ya/yi follow anything other than a predicate in Old Sinhala, and specifically in this class of construction, it always appears (when it appears, since it is optional) following the nominalised verb and never following a focussed constituent, unlike in later Sinhala.

The identification of the early and classical Sinhala ya/yi/i as an agreement element is clear in its pre-Colloquial distribution, where it appears as one out of a set of agreement elements, cp. (11a), (11b), (11e), (11f) with (11c), (11d).10

The function of ya as an agreement element rather than a focus marker is clear also in examples like (12), which contains an explicit focus marker nu immediately following the focussed phrase, with ya immediately following the nominalised verb visi.

Quite frequently in the Sigiri graffiti texts, when impersonal-type nominalisations occur, there is no apparent focussed element. Example (12) is an unusually clear example of a focussed constituent (as indicated by use of the focussing particle nu) occurring with an impersonal.

As indicated above, (12) also illustrates the role of ya as an agreement clitic which appears immediately following the impersonal nominalised verb. Also notable in (12) is the (at this stage optional) dislocation of the focussed element to clause right-edge, a common feature (as discussed above) of later Sinhala. We do also find examples of impersonal nominalised verbs occurring when a focussed element occupies the clause, but without dislocation of the focussed element and without any overt form of ‘to be’ or agreement clitic, as in (13).11

Thus, as discussed in the next section, Sinhala focus concord constructions seem to have originated as a possible usage of the impersonal construction. This construction involved the use of a nominalised verb in a typically monoclausal/non-clefting structure which was compatible with but not specialised for focus, and from there developed into a “true” focus-associated cleft construction, where the nominalised participial clause enters into a predication relationship with the focussed element. In the modern period this construction again evolved into a non-predicative structure, but with no true nominalised verb.

Though focus concord constructions in earlier Sinhala clearly involve nominalised verbs, this is no longer the case in Modern Colloquial Sinhala. In modern Sinhala, true verbal nominalisation involves either the “gerund” form of the verb in -iimə or -illə, (14a), or, more commonly, the use of the “adjectival” form of the verb (generally followed by ekə if there is no overt noun; where ekə is etymologically the inanimate numeral ‘one’), (14b) – both of which are clearly morphologically distinct from the verbal “-E” form appearing in focus concord constructions.

In this, modern Sinhala focus concord constructions stand in contrast to Dravidian focus concord constructions. In Tamil, for instance, the focus concord verbal form in (15a) is identical to clear verbal nominalisation as found in instances like (15b).12 The same form found in the focussing (15a) appears also in the formation of the action nominal in the nominalised clause of (15b) below.

In Old and Classical Sinhala, on the other hand, it is clear that the focus concord musubi verbal forms are in fact nominalised verbs, and in Old Sinhala, as discussed earlier, we find a sort of “proto focus-concord” form. Since the construction in Old Sinhala is in fact an impersonal, the use of a nominalised verb is compatible with the presence of a focussed element, but does not require one. That is, in Old Sinhala the employment of this type of verbal form does not automatically correlate with the presence of a focussed element, as illustrated by examples like (9), repeated below as (16).

In Old Sinhala, the position of the agreement clitic or form of ‘to be’ indicates that this “impersonal” nominalised verb construction – with a nominalised main verb and an (optional) agent in the genitive case – seems to generally be monoclausal, being generically paraphraseable as ‘there is X-ing (by Z)’. Examples like (12), however, where a clearly focussed element stands in a dislocated position would appear to invite the possibility of a reanalysis which seems to have fully taken place by the Classical Sinhala period. That is, in this later stage the focus structure would be roughly paraphraseable as ‘the X-ing is Y’.

Thus, by the Classical Sinhala period the “impersonal” construction of Old Sinhala appears to have been re-analysed as predicative clefting constructions, where the focussed element acts as a predicate for which the nominalised participial is an argument.

That such constructions involve predicative clefting in Classical Sinhala is suggested not only by the obligatoriness of an agreement clitic or form of ‘to be’, but also by the obligatory overt dislocation of the focussed element to either the clause left- or right-edge. And the shift to a predicative construction is suggested further by the fact that, unlike in Old Sinhala, the agreement element appears not after the nominalised verb but obligatorily after the focussed element. These features are all observeable in (7b), repeated below as (17), with partial tree representations of the relevant pieces in (18a) and (18b) (showing the two movement steps), with the focussed phrase “normalised” to the right-edge.

First, as shown in (18a), the focussed element moves to a (VP-internal) focus position, adjoined to VP. Then, as in (18b), the nominalised clause moves to SpecTP, where it can enter into a predication relationship with the focussed element (=‘the giving of alms by me is for them’).

In contrast, the Old Sinhala equivalent of (17) would have the structure of the constructed example in (19), with a partial tree of the relevant pieces given in (20).14

Here the V head in which yæ appears, if vocalised, remains associated (and right-adjacent to) the nominalised participle, with optional movement of the focussed element to a focus-associated position. But this V does not serve to create a new predicate as it were, unlike in (18); ovun san̆dahā ‘for them’ does not stand in a predicative relationship with the nominalised mā dan denne (lit.) ‘a giving of alms by me’, but is rather an extracted focussed adjunct of the NP.

The comparison between (17) and (19) illustrates the shift from the non-predicative “impersonal” construction in (19), compatible with the focussing of a particular constituent but not inexorably tied to this type of highlighting function, to the true predicative “cleft” construction of (17).

MLS focus-concord constructions closely resemble those of the Classical period, except that the dislocation of focus is restricted to the right edge. However, in MCS we find the old agreement clitic -y (<ya) being re-analysed simply as a focus marker.15 Modern Sinhala also shows that the focus concord verbal form became distinct from verbal nominalisation.16 With these changes, the dependency between the focussed element and the verbal form has become something like agreement (i.e. the verb takes a special morphological form when an exhaustive-focus element originated within its scope). Thus the MCS sentence in (21) would be associated with the partial tree structure given in (22).

Thus the MCS focus construction is similar in some ways to the early structure we find in OS, with the difference that no nominalisation is involved.

That -y (<ya/yi) is no longer a (verbal) agreement element in MCS is also attested by the difference in the structure of clauses with non-verbal predicates in MCS as opposed to earlier varieties. In MCS, there is no overt realisation of agreement on verbal forms; and in clauses with non-verbal predicates no copula or agreement appears whatsoever:

Whereas, in MLS, clauses with non-verbal predicates require some sort of overt agreement element, either an agreement clitic (of which yi/ya is a member), as in (24), or a form of ‘to be’, as in (25).

Note also in Modern Colloquial Sinhala that the focus construction now behaves identically to non-focussed constructions with respect to the case-marking of the subject (as opposed to the accusative subjects of literary Sinhala “musubi” verbs). The loss of overt subject-agreement morphology on the finite verb in Modern Colloquial Sinhala may be one reason for this difference.18

Table 1 lays out the properties of focus concord structures in these four stages of Sinhala, including the location of the agreement element (i.e. forms of ‘to be’ or agreement clitics, when present), the case of the subject of the focus-concord clause, the dislocation of the focussed element, and the status of the verb which appears in the focus-concord clause (as identical or non-identical with a nominalised verb). Note that the “accusative” of Classical and Literary Sinhala is the erstwhile genitive (see fn. 19).

The next section provides a brief overview of the properties of Q-particles in Sinhala, before turning to the comparison with Dravidian in Section 6.

This section examines constructions in Sinhala containing the particle də, focussing on Modern Colloquial Sinhala, but including some examples from earlier varieties as well, in order to provide an overview of the full distribution of the interrogative focus-association particle, and how it interacts with the “musubi” verbal form. I first begin by examining the historical predecessors of Sinhala də.

The Sinhala particle da/də ultimately derives from Old Indo-Aryan utā́ho, which is made up of two particles, ā́ho and uta.21 In Sanskrit (as a representative Old Indo-Aryan language), particularly in the late Vedic period, these two particles appear in alternative questions, as (26).

The particle ā́ho is of unclear origin in this usage. Uta appears in early Vedic Sanskrit with the sense ‘and’ (Klein 1978; see also Klein 1974), as in (27).

It is from this utá A B C … ā́ho X Y Z construction that the form utā́ho seems to derive. Since utá may appear at the front of an interrogative clause (functioning as a sort of interrogative marker), it could also immediately precede the ā́ho standing at the front of the second part of a disjunction, with the application of sandhi becoming utā́ho. And it is in this form (utāho), that it appears in Classical Sanskrit, as in (28) below.

In later Pali, we find the cognate form udāhu, as in (29).

By the earliest extant Sinhala texts containing questions, it seems that the descendent of utā́ho/udāhu, i.e. the Sinhala particle da/də, has been generalised from alternative questions and has begun to appear (though apparently optionally) in wh- and yes/no-questions as well.

Sinhala da (later də in MCS) appears obligatorily in interrogatives in modern Sinhala (both literary and colloquial), triggering the use of the focus concord E-verbal form. Additionally, it is always the case in literary, and often in colloquial, that the wh-phrase is dislocated to the right of the verb. Da/də, as do other focus-associated particles, triggers the focus concord E-verbal form whenever it originates inside of the c-command domain of the verb (regardless of whether later dislocating/clefting operations later move it outside).

In “neutral” yes/no-questions (i.e. where there is no focussed element) the particle də originates outside of the c-command domain of verb, and the verb therefore occurs in the neutral (“A”) form, as in (32a). Where a yes/no-question contains a focussed element, the particle də follows it, as in (32b) (and the focussed element together with də may optionally be dislocated to the right clause-edge).

In alternative questions, each disjunct is followed by an instance of the particle də, and the verb appears in the E-focussing form, as shown in (33).

In Modern Colloquial Sinhala, but not in earlier varieties, də can combine with a wh-word to form an indefinite (cf. Slade 2015). N.b.: the verb never appears in E-focussed form with wh-indefinites, as shown in (34).

In non-interrogative disjunctions, a morphologically distinct Q-particle hari (hō in literary Sinhala) appears rather than də.23 Hari/hō does not trigger focus concord E-verbal forms, as shown in example (35).

In Modern Literary Sinhala, and earlier forms, we find similar disjunctive structures, but with the particle hō rather than hari, as shown in (36).

Both literary and colloquial Sinhala can form wh-based indefinites by combining a wh-word with the particle hari, in MCS, as in (37), or hō, in MLS, as in (38).

In NPI contexts vat is employed instead of hari/hō, both in disjunctive contexts as shown in example (39a), and in the case of wh-based indefinites, as in (39b).24

The Sinhala Q-particle də/da thus did not start out with inherently focal semantics, though it acquired association with focus by the period of modern literary language. In Modern Colloquial Sinhala, we find that – alongside its use as a Q-particle – də is also used in the formation of a certain class of indefinites and thus in these cases does not trigger verbal concord “musubi” forms, and therefore the particle də itself is not always focus-associated in the Modern Colloquial language.

In Old and Classical Sinhala, questions do not require the kakari-type particle da, and da does not automatically trigger the use of a nominalised “musubi” verb. In Modern Literary Sinhala, on the other hand, questions do require the appearance of da, and da always triggers the use of a nominalised “musubi” verb (when da appears in its domain). While in Modern Colloquial Sinhala, questions do require də, but də does not trigger musubi-type focus-concord verb forms when it occurs in the formation of an indefinite (as discussed in the previous paragraph).

Table 2 shows the patterning of association of də/da and hari/hō with verbal focus forms.

In the next section, I discuss focus concord constructions and associated particles in the South Dravidian languages Tamil and Malayalam.

Focussing/cleft constructions in Dravidian involve a nominalised verb, often with dislocation of the focussed element to the right-edge, and are thus reminiscent of focus concord constructions in Sinhala. In this section I focus mainly on the Dravidian languages Tamil and Malayalam, with comparison to Sinhala.

Compare the Tamil “neutral” (no focus) example (40a) with example (40b), where the verb shifts from a finite form to a nominalised form (again, similar to earlier Sinhala), with optional marking of focus on the focussed element via a particle -tāṉ.

As in Sinhala, focus constructions in Tamil can often involve “clefting” of the focussed element, dislocating it to the right-edge of the clause, as shown in example (41).

The Tamil focus particle -tāṉ can be replaced with another focus particle such as -ē or with the interrogative particle -ā in focussing yes/no-questions like (42).

The Tamil interrogative particle -ā does not obligatorily trigger clefting or the use of a nominalised (“musubi”) verb form, as shown by the grammaticality of (43).

Differing from the Sinhala pattern, wh-questions in Tamil do not appear with an interrogative particle, and they do not usually employ a nominalised “musubi” verb form; an example of a typical Tamil wh-question is given in (44).

Tamil stands in contrast with Malayalam on this point. In Malayalam, wh-questions normally appear using a focus concord verbal form, as in (45).25

The Tamil particle -ō appears in the formation of non-interrogative disjunctions, as shown in (46a), and in the formation of wh-based indefinites, as shown in (46b), both of these being reminiscent of the behaviour of Sinhala hari/hō, in the appearance of the particle on both disjuncts in (46a) as well as the non-triggering of a focus-concord form.

In modern Malayalam, the particle -ō also occurs in these two contexts (non-interrogative disjunctions and in the formation of indefinites), but additionally occurs also in the formation of yes/no-questions, in place of Tamil’s -ā; an example is given in (47).

In Old Malayalam, -ō additionally occurs in the formation of wh-questions (making it more closely resemble Sinhala), as in example (48).

Old Tamil26 resembles Malayalam in not possessing a distinction between -ā and -ō, and further resembles Old Malayalam in (optionally) employing -ō in wh-questions as in (49a), and in yes/no-questions rather than -ā (in variation with -kol, and the apparently composite form kollō) as in (49b).27

Note that in modern Malayalam, unlike in modern Tamil, a form of ‘to be’ is generally required to follow the focussed element, as in (50).

While -ō appears in the formation of wh-based indefinites in Old Malayalam and in the modern forms of both Malayalam and Tamil, such forms do NOT occur in Old Tamil (p.c. Thomas Lehmann), and so the appearance of these forms in later stages of Dravidian represents an extension, similar to what we see in Sinhala, where hō appears in the formation of wh-indefinites only from Modern Literary Sinhala onwards, and da/də does not appear in this environment until the Modern Colloquial stage.

Thus it can be seen that in early South Dravidian the “clefting” focus-concord construction remains independent of the use of focus particles (including Q-particles) and vice versa. This remains the case in modern Tamil, but not in modern Malayalam, where Q-particles do trigger the use of focus-concord verbal forms. Both early and modern Tamil then have a somewhat similar state of affairs as in Old Sinhala, as discussed above in Section 3, though the clause structure differs (i.e. plain “impersonal”/“existential” nominalisation in Old Sinhala, against predicative clefts in Dravidian) as shown by the positioning of agreement elements (i.e. forms of ‘to be’ or agreement clitics).

Thus – unlike in varieties of Sinhala from the Classical period onwards – Q-particles in early South Dravidian as well as in modern Tamil do not trigger nominalised “musubi” verb form or clefting of the focussed element. Interestingly, modern Malayalam is somewhat more similar to post-Old Sinhala in that focus particles do trigger the focus concord verbal forms. And as in Classical & Literary Sinhala, Dravidian clefting/focus constructions (at all stages) appear to be copular/predicative.

Thus, in recapitulation and summary, the development of focus concord constructions in Sinhala and Dravidian can be described as follows.

In Old Sinhala, we find nominalised verbal forms used in what might be described as “impersonal” constructions, of the type ‘there was X-ing (by Y)’, as in (13), repeated below as (51).

Such structures were compatible with focussing, and we do sometimes find clear instances of focussed elements used in this nominalised impersonal construction in Old Sinhala, as in example (12), repeated below as (52).

In terms of the structure, the Old Sinhala nominalised impersonals do not involve any copular clefting; where a form of ‘to be’ is overt, it always immediately follows a nominalised impersonal verb, and thus it is not “copular” in the sense of being a clause linker (as the function of such verbs in focus-concord constructions might be described in later Classical and Modern Literary Sinhala), but can be seen as part of the deverbalisation of the impersonal.

In Classical Sinhala and Modern Literary Sinhala such constructions do in fact appear to involve predicative clefting, perhaps under Dravidian influence. At this stage focus-concord constructions thus take the form of a nominalised clause joined to the remainder by an obligatory copula or agreement clitic, as in (7a), repeated below as (53).

Modern Literary Sinhala is quite similar to Classical Sinhala, with the exception that focus concord constructions show less variation and can be regarded as more frozen,28 since the focussed element obligatorily dislocates to the right of the verb.

By the period of Modern Colloquial Sinhala, the agreement element is reanalysed as a focus marker, and so focus concord constructions in Modern Colloquial Sinhala once again become non-predicative (though in a rather different fashion than their Old Sinhala precursors). Overt dislocation is thus optional, allowing for both (54a) and (54b) (repeated from above (1b), (1c), respectively):

The Modern Colloquial Sinhala focus concord construction is no longer “copular” (predicative clefting), thus differing from its Literary and Classical counterparts. In Modern Literary Sinhala, ya or a form of ‘to be’ must appear in focus constructions, while in Modern Colloquial Sinhala, -y (<ya) is optional, and can be substituted by a different focus particle (or no post-focus element may occur at all).

The Sinhala Q-particle də/da did not start out with inherently focal semantics, though it acquired association with focus by the period of the Modern Literary language. In Modern Colloquial Sinhala, we find that — alongside its use as a Q-particle – də is also used in the formation of a certain class of indefinites and thus in these cases does not trigger verbal concord “musubi” forms, and therefore the particle də itself is not always focus-associated in the Modern Colloquial language.

In Old and Classical Sinhala, questions do not require the kakari-type particle da, and da does not automatically trigger the use of a nominalised “musubi” verb. In Modern Literary Sinhala, on the other hand, questions do require the appearance of da, and da always triggers the use of a nominalised “musubi” verb (when da appears in its domain). While in Modern Colloquial Sinhala, questions do require də, but də does not trigger musubi-type focus-concord verb forms when it occurs in the formation of an indefinite (as discussed in the previous paragraph).

Table 3 presents a concise summary of the properties of particles in various stages of Sinhala and South Dravidian.

As discussed above, the distribution of “kakari”-type Q-particles also differs between languages, and between language stages, as summarised in Table 4.29

The Modern Colloquial Sinhala focus concord construction is no longer “copular” (predicative clefting), thus differing from its Literary and Classical counterparts. In Modern Literary Sinhala, ya or a form of ‘to be’ must appear in focus constructions, while in Modern Colloquial Sinhala, -y (<ya) is optional, and can be substituted by a different focus particle (or no post-focus element may occur at all).

While in Japanese focus concord constructions appear to have originated as focus-associated, the construction which eventually becomes a focus concord construction in Sinhala did not originally obligatorily associate with focus. Further, particles (including the Q-particle də/da) which end up triggering the appearance of a verbal focus concord form in later Sinhala did not originally require the use of a special verbal form. In Dravidian, both contemporary and earlier, Q-particles do not force the appearance of the special “clefting” (focus concord) construction; focus particles do not trigger the focus concord/“musubi” form either in early South Dravidian nor in Modern Tamil, but they do in Modern Malayalam.

Sinhala and Dravidian Q-particles stand in contrast to Japanese ka, whose extant forms are all focus-associated, up until the 17th c., when it becomes a question operator without any overt focus-association, i.e. it no longer triggers musubi focus-concord verbal forms (Serafim & Shinzato 2000; Aldridge 2009; Mitrović 2014a). Sinhala da/də appears to show the reverse development, originating as a question-associated particle which later becomes focus-associated. The Dravidian Q-particles -ā and -ō, on the other hand, never appear to become obligatorily focus-associated.

Intriguingly, although the Sinhala, Dravidian, and Japanese focus concord constructions overlap in some of their features, Sinhala and Japanese appear to show largely opposite paths of development in terms of the association of “musubi”-type particles with questions and focus, but in both languages focus-concord constructions develop out of earlier predicative “clefting” focus constructions.

Though the development of Sinhala focus-concord constructions thus differs on many points from that observed in Japanese, there are some similarities, especially in the later developments. Sinhala focus-concord constructions have been reanalysed as non-predicative focus-marking in the Modern Colloquial language. So too, while Japanese kakarimusubi constructions seem to have originally been predicative clefts, later examples are not necessarily analysable as such (Akiba 1978: 77; cp. Harris & Campbell 1995: 161–162), in part due to the fact that some kakari particles like zo no longer require the nominal form and the fact that other kakari particles, like nan, have more or less disappeared (Sansom 1928: 266). Parallels exist elsewhere: present-day monoclausal focus constructions in Somali which developed from earlier biclausal constructions (Hetzron 1974; Antinucci & Publielli 1984); and Breton shows a similar development (Harris & Campbell 1995: 155–158).

The earlier development of the “monoclausal” impersonal construction of Old Sinhala – which was not directly tied to the presence of focus, but was certainly compatible with it – to the “biclausal” predicative clefting focus-concord construction of Classical Sinhala may seem on the face of it an unusual change, as it involves the innovation of a complex structure out of a simpler one. However, it too finds parallels elsewhere. Harris & Campbell (1995: 311–312) provide an example of a biclausal structure developing from an earlier monoclausal structure in Udi, a Lezgic language. Aside from Udi, Lezgic languages lack finite relative clauses, as do sister languages in the other branches of the Northeast Caucasian language family. In Udi, in very recent times, finite relative clauses have developed, and out of an earlier monoclausal structure:

As discussed by Harris & Campbell, the deverbal adjectival modifier gölöšpi ‘danced’, derived from the verbal gölöšp- ‘dance’, could be re-analysed as a finite verb (since Udi permits clauses consisting of a surface structure containing just a finite verb). So the AdjP gölöšpi ‘danced’ could be reanalysed as a clausal ‘(s/he) danced’, thus deriving a biclausal structure from a monoclausal one. As Harris & Campbell (1995: 310–311) say, ‘[n]on-finite verbal forms – deverbal nouns and adjectives – have an inherent dual nature, which can lead naturally to dual analysis […] being at once substantival and (de)verbal, they have the potential for being diachronically reanalyzed as having a complex initial structure’.

The situation in Sinhala is slightly different. Here we do not find the reanalysis of a deverbal form by itself, but rather the reanalysis targets the combination of a deverbal form with an agreement clitic or form of ‘to be’. That is, the reanalysis involves whether the agreement element (actually optional in Old Sinhala) is associated with the deverbal nominal simply as an agreement element, as it is in Old Sinhala, or whether it acts as a sort of predicative linker, as it does in Classical Sinhala. This reanalysis takes advantage of a somewhat different inherent duality, namely whether the agreement clitics or forms of ‘to be’ function purely as expressions of morphosyntactic features (as in “impersonal” nomalised constructions in Old Sinhala), or as predicative/equatative linkers between elements (as in the “cleft” focus constructions in Classical Sinhala).