3.1 Cartography of the CP-domain

Prior to discussing how focus particle movement takes place in the focus concord constructions, let us delineate some assumptions about Sinhala clause structure. First, as discussed by Kishimoto (2005), Sinhala can be regarded as a double CP (or CP-recursion) language, where the complementizer kiyəla ‘that’ appears alongside an interrogative particle in embedded interrogative clauses, as exemplified in (9).

In (9), the interrogative də-næddə ‘whether’ appears to the left of kiyəla ‘that’.⁵ Given that an ordinary complementizer appears alongside an interrogative marker, it is plausible to state that in Sinhala, the CP domain includes two complementizer projections; one is filled by an ordinary complementizer, and another, by an interrogative complementizer (cf. Bhatt & Yoon 1991; Authier 1992; Browning 1996; and others).

With regard to the CP structure, I essentially follow Rizzi’s (1997; 2004) proposal that the CP domain consists of at least four distinct projections which take on different discourse-syntactic functions, as in ForceP-TopP-FocP-TopP-FinP (see also Coniglio and Zegrean 2012).⁶ Note, however, that Rizzi advances his proposal based on the facts of Italian, which is not a double CP language. On Rizzi’s proposal, ForceP, specifying its clause type as declarative, interrogative, etc., is located on the top in the CP system, but as seen above, in Sinhala, the ordinary complementizer kiyəla—which appears regardless of whether the embedded clause is interrogative or declarative—is positioned to the right of də-næddə ‘whether’ or a clause-final focus particle (e.g. də and tamay). This fact suggests that there is a projection accommodating kiyəla projected above ForceP.

Given that ForceP is a locus where clause type is determined, and further, that the ordinary complementizer kiyəla licenses complementation regardless of clause type, I postulate that a complementation-licensing projection, referred to as ComplP, is projected above ForceP, as illustrated in (10) (cf. Kishimoto 2005).

I assume that the type of complement clause is determined according to whether or not an interrogative particle (i.e. də/də-næddə) fills ForceP. If a complement clause has ForceP filled by an interrogative particle, it is identified as an interrogative clause; if it remains unfilled by an interrogative particle, it is identified as a non-interrogative clause. (Note that the upper verb imposes a selectional restriction on ForceP, while ComplP (indicating complementation) is transparent to this selection.)

In Sinhala, the function of CP is split into two projections: ComplP licenses the embedding of a complement clause and ForceP determines the illocutionary force of the clause. The complementizer kiyəla is used to license complementation, but its deletion is possible with a limited set of verbs including kiyənəwa ‘say’ and ahanəwa ‘ask’ (but not dannəwa ‘know’) when the complement clause is adjacent to the verb, as (11) illustrates.

The complementizer kiyəla cannot be elided when the clause is moved to the front, however, as shown in (12).

The complementizer kiyəla can be elided only when complementation information can be recovered with an adjacency relation to the verbs. (12b) suggests that də-næddə occupying ForceP cannot mark the status of the complement clause (and the same holds true of the focus particles də and tamay). The complementizer kiyəla behaves in the same way as the English complementizer that, which can be omitted when it appears adjacent to verbs like say, as in John said (that) he would come. The distribution of complementizer deletion shown in (11) and (12) would not be expected unless kiyəla has a syntactic function of licensing complementation, just like the English complementizer that. Furthermore, a comparison of (11b) and (12b) suggests that kiyəla and də-næddə are separate lexical heads, and thus, it can be hypothesized that ComplP, occupied by kiyəla, is projected above ForceP, filled by də-næddə.⁷

With the articulated clause structure (10) in mind, let us now turn to the discussion of how focus particles behave in Sinhala focus concord constructions. Sinhala has a number of focus particles with different focus-related meanings, and də is distinguished from others, in that it has an interrogative import.

The difference in the kind of quantificational force encoded in focus particles can be readily detected by placing them in the complement clause selected by ahanəwa ‘ask’, which takes an interrogative clause as its complement.

As shown in (14), də makes the clause interrogative, and hence the clause comprising də can be a complement clause of the verb ahanəwa ‘ask’, but the non-interrogative clauses comprising other focus particles cannot.

With regard to the question of where clause-final focus particles appear in clause structure, there are two possible positions that can be filled by focus particles.

I suggest that focus particles serve as operators in the CP layer, and that a non-interrogative focus marker like tamay appears in FocP (indicated by α), while the particle də is licensed in ForceP (indicated by β) as an interrogative operator.

To be a little more concrete, the general discourse function of a non-interrogative focus particle like tamay is to encode non-background information on the constituent to which it attaches. Given that the focus particle appearing in clause-internal position is associated with focus, I suggest that it is raised to FocP before the derivation goes into the C-I interface, as depicted in (16).

The interrogative focus particle də has an additional function of specifying the clause as “interrogative” (Chen 1991). This suggests that də would be related to ForceP, alongside FocP. Thus, if the focus marker də starts out in clause-internal position, it is moved to ForceP via FocP, owing to its function of specifying the illocutionary force of the clause (Rizzi 1997; 2004), as (17) illustrates.

When an interrogative operator resides in ForceP, as in (17), the clause is interpreted as interrogative. In cases where no operator specifying illocutionary force resides in ForceP, the clause is interpreted as declarative/non-interrogative as a default interpretation (see e.g. Bošković 1997).

3.2 Long distance A’-extraction

For the purpose of illustrating that focus particles occupy two distinct structural positions in the CP domain, which can be distinguished according to whether or not they carry an interrogative meaning, this section shows that the patterns of wh-island effects induced by long distance extraction of DPs differ depending on the type of clause.

Prior to discussing the facts of wh-island violations found in focus concord constructions, observe that pseudo-clefting is a syntactic operation that places a constituent in the focus position located to the right of the predicate, as in (18b).

The pseudo-cleft sentence in (18b) is derived from (18a) by moving ee potə ‘that book’ to the right end of the clause.⁸

In Sinhala, an asymmetry is observed with regard to long distance pseudo-clefting, depending on the type of embedded clause. The verb dannəwa ‘know’ takes either a declarative or interrogative clause as its complement. A DP can be extracted from a non-interrogative complement selected by dannəwa via pseudo-clefting without causing any island effect, as in (19).

By contrast, when the verb dannəwa takes a də-næddə ‘whether’ clause, the extraction of a DP from the embedded clause induces a violation of the wh-island constraint.

Given that the sole structural difference between (19) and (20) lies in the presence or absence of də-næddə, it is reasonable to state that the wh-complementizer placed at the end of the embedded clause is responsible for the wh-island effect in (20).

Let us now turn to the discussion of focus concord constructions. When the verb dannəwa ‘know’ takes a wh-interrogative complement, a difference in acceptability is observed with regard to the long distance extraction of a DP like ee potə ‘that book’, as shown by the pair of the sentences in (21a) and (21b).

In (21), the presence or absence of a wh-island effect depends on the position of the interrogative focus particle də. When də appears in clause-final position, a wh-island effect is induced, as in (21a). When it appears in clause-internal position, no wh-island effect is observed, as in (21b).

When the verb dannəwa ‘know’ takes a yes-no interrogative clause as its complement, the interrogative focus particle də induces a wh-island effect regardless of whether it is placed in clause-internal or clause-final position, as shown in (22).

The yes-no interrogative clauses in (22) stand in contrast to the wh-interrogative clauses in (21), since a wh-island effect is observed even when the focus particle də is placed in clause-internal position.

Furthermore, if the verb dannəwa ‘know’ takes a declarative complement, no wh-island effect is induced regardless of the position of the focus particle.

The examples in (23) show that a declarative particle like tamay does not block the long distance extraction of DP even when it is placed at the end of the embedded clause.⁹

In addition, observe that the same contrast in acceptability that is found in (19) and (20) is observed for the long distance “leftward” movement of DP out of the complement clause, as shown in (24).

Example (24a), where ee potə ‘that book’ has been scrambled to the sentence front, is well-formed, but (24b) displays a wh-island effect, because it has been extracted from the də-næddə ‘whether’ clause.

It is worth noting that long distance scrambling gives rise to exactly the same paradigms as observed for long distance pseudo-cleft extraction in the focus-concord constructions. The following examples illustrate the point.

The facts suggest that wh-island effects are caused by not only long distance pseudo-clefting, but also other instances of long distance A’-movement (including scrambling).

The data (21) through (23) illustrate that the patterns of the wh-island effects incurred by long distance pseudo-cleft extraction differ, depending on the type of focus concord construction involved, as summarized in Table 1.

Long distance pseudo-cleft extraction is consistently not blocked by non-interrogative focus constructions, but in yes-no interrogative focus constructions, the particle də does yield a wh-island effect regardless of its surface position. In wh-interrogative focus constructions, the particle də placed in clause-final position gives rise to the effect of a wh-island violation for long distance pseudo-cleft extraction, but this effect does not arise when the particle is placed in clause-internal position.

Note that clause-internal focus particles behave like phrasal elements. To make this point, observe first that it is not possible to extract DPs from syntactic islands via pseudo-clefting. The examples in (28) represent cases involving extraction from a relative clause and a noun complement clause.¹⁰

The examples illustrate that the relative and noun complement clauses constitute islands for A’-extraction, and that when a DP is extracted from the complex DP islands via pseudo-clefting, the sentences are rendered unacceptable.

In focus concord constructions, the same island effects that are observed in (28) obtain when a declarative focus particle like tamay is placed inside the complex DP islands, as shown in (29).

In (29), the focus particle is attached to a DP inside the relative and the noun complement clauses, and the matrix predicate has the -e marking, indicating that the focused constituent takes matrix scope. No visible extraction takes place from the syntactic islands in (29), as opposed to (28). Nevertheless, the same island effects that are obtained for long distance pseudo-cleft extraction in (28) are observed in the focus concord constructions in (29).

While the sentences in (29) display the complex DP island effects, no such island effect emerges in (30), where the focus particle tamay is attached to the right edge of the islands.

In (30), the entire DPs to which the particle tamay is attached receive focus, i.e. the entire islands count as focused constituents. They take matrix scope, as the main predicate has the -e marking. In (30), no island effects are observed.

The same patterns that are observed in non-interrogative focus constructions are found in wh-interrogative focus concord constructions. In (31), the interrogative particle də appears inside the syntactic islands, and the matrix predicate assumes the -e marking; the sentences give rise to island effects.

No island effects are obtained, however, when the interrogative particle appears at the right margin of the syntactic islands.

In (32), the interrogative particle is not embedded in the syntactic islands, and the sentences are well-formed direct wh-questions. A comparison of (31) and (32) shows that the interrogative particle də is responsible for the island effects observed in (31), even though no visible movement takes place (see Kishimoto 1992; 2005; Hagstrom 1998).

The yes-no interrogative focus concord constructions with the interrogative particle də show exactly the same distribution. Island effects are obtained in the examples in (33), where the interrogative particle is embedded inside the syntactic islands.

By contrast, when the interrogative particle is placed to the right of the syntactic islands, no island effects are observed.

Again, the focus particle is not embedded in the syntactic islands in (34). The examples are well-formed as direct yes-no questions.

In the examples (29) through (34), the matrix predicates bear the -e marking, indicating that the focused constituents take matrix scope. When a focus particle is embedded in syntactic islands, island violations are incurred. It is easy to see here that the island effects observed in (29), (31), and (33) arise by virtue of the focus particle undergoing invisible A’-movement out of the syntactic islands, as illustrated in (35a).

In (35a), the focus particle (FP) is extracted from an island, and this particle movement to its scope position results in a violation of the island constraint. In (35b), by contrast, the focus particle (FP) occurs at the right margin of the island; in this case, no island violation is caused even if the focus particle undergoes movement to its scope position. Importantly, in the focus concord constructions, the island effects are observed even if no visible extraction of constituents is involved.

Focus particles display island effects when embedded in syntactic islands, but not when placed at the edge of the islands. The patterns that we observe for focus concord constructions are parallel to those found in A’-movement of DPs. Therefore, it can be stated that focus particles behave as operators undergoing phrasal A’-movement (suggesting that they are non-projecting heads behaving as phrases syntactically; see Chomsky 1995), and that the island effects are caused by A’-movement of focus particles out of syntactic islands. Given that particles are placed at the clause end when they do not delimit focus, I assume that operator elements (including null operators) appear in clause-final position, i.e. the operator positions are located at the right periphery of the clause.

In the next section, I suggest that the distribution of the wh-island effects found in (21), (22), and (23) can be accounted for in terms of the structural position that the focus particles occupy in narrow syntax.

3.3 Focus particle movement in narrow syntax and LF

As discussed previously, the distribution of wh-island effects that arise when a DP is extracted from the complement clause comprising the interrogative focus marker də differs according to whether the complement is a wh-interrogative or a yes-no interrogative clause. In this section, I argue that in the wh-interrogative clause (selected by dannəwa ‘know), an interrogative particle can appear clause-internally, because the Force head can have a weak operator feature, which motivates LF focus particle movement. On the other hand, I suggest that in a yes-no interrogative clause, overt syntactic focus particle movement is invoked on clause-internal interrogative focus marker də due to a strong operator feature on the Force head, but that superficially, the particle appears in clause-internal position owing to the pronunciation of the initial copy of the operator chain, which is necessary to demarcate the syntactic division of presupposition and focus. It is also suggested that non-interrogative focus concord constructions do not exhibit wh-island effects for long distance pseudo-cleft extraction regardless of the surface position of their focus particles, because non-interrogative focus particles fill FocP, and do not move up to ForceP in narrow syntax (NS).

To be concrete, recall that long distance pseudo-cleft extraction is fine with no wh-island effect invoked in (36a) (=(19)), while the same long distance movement incurs a wh-island violation in (36b) (=(20)).

For ease of reference, I assume here that the pseudo-cleft construction involves movement of a null operator OP, which is linked to the DP in cleft-focus position via predication. The difference in acceptability between (36a) and (36b) can be attributed to the presence or absence of the complementizer də-næddə in ForceP.¹¹ In the present analysis, the projection accommodating də-næddə is ForceP. (Recall that the complementizer kiyəla ‘that’ appearing in ComplP is invariant regardless of clause type). Thus, ForceP can be assumed to constitute an island.¹² This suggests that long distance movement needs to proceed through ForceP if it is to be extracted from the clause.

Long distance extraction of a null operator is possible when the embedded clause does not include də-næddə, as it can move through ForceP in the embedded clause, as (37a) illustrates. By contrast, when də-næddə appears in the clause, a wh-island effect is observed for long distance pseudo-cleft extraction. The long distance DP extraction is precluded in (37b) due to the presence of də-næddə, which appears in the embedded ForceP.

If ForceP serves as an intermediate landing site for long distance A’-movement, it is possible to assess whether ForceP is filled by an interrogative focus particle in narrow syntax. In the first place, in wh-interrogative focus concord constructions, a wh-island effect arises when the focus particle də appears at the clause end, as in (38a)(=(21a)), but no such effect is caused by də occupying a clause-internal position, as in (38b)(=(21b)).

The wh-island effect observed in (38a) suggests that the clause-final particle də occupies ForceP, blocking the long distance pseudo-clefting.

If ForceP is filled by the focus particle də, long distance pseudo-cleft extraction cannot have access to it. The wh-island effect is caused in (38a) by illicit long distance movement, which skips ForceP, as illustrated in (39). The wh-phrase, which perforce receives focus, needs to be paired with də, which delimits focus, so I assume, with Kishimoto (2005), that də is first merged in a clause-internal position, most typically in a position contiguous with the wh-phrase. Since the interrogative particle has the function of specifying the scope of its associated focus constituent, this particle appears in clause-final position in (38a) by virtue of undergoing overt syntactic movement to ForceP by way of FocP.

On the other hand, in (38b), where an interrogative focus particle appears in clause-medial position, no wh-island effect is observed. This suggests that the focus particle does not appear in ForceP in narrow syntax. Nevertheless, a clause-internal focus particle incurs island violations when it is placed inside syntactic islands, as in (40)(=(31a)).

In light of the island violation observed in (40), it is fair to state that in wh-interrogative focus concord constructions like (38a), the focus particle də is first placed in the focus-delimiting position, and then moves to the scope-marking position in narrow syntax, as depicted in (41).

For a clause-internal focus particle, there are two possibilities for the timing of its movement, which causes island violations when it is embedded inside syntactic islands.

Traditionally, it has been assumed that only overt syntactic movement incurs island violations (e.g. Chomsky 1981; Huang 1982), but there are also claims that island violations could be caused by LF movement (e.g. Choe 1987; Nishigauchi 1990). Thus, one can claim that the island violations in Sinhala focus concord constructions where focus particles are placed inside syntactic islands are incurred by their LF movement.

In the LF movement analysis, a focus particle placed in clause-internal position remains unmoved in narrow syntax, and hence its copy does not appear in the CP domain. Since the particle is moved to the CP domain at LF, LF movement of the particle is responsible for the island effects.

There are also some recent attempts to reduce the distinction of overt syntactic versus LF movement to the question of whether the head or the tail member of a chain created by movement is pronounced (see e.g. Brody 1995; Groat & O’Neil 1996; Pesetsky 1998; Bobaljik 2002). Thus, the analysis taking the island violations to be caused by overt syntactic movement, as in (43), is also conceivable.

In the overt syntactic movement analysis, the particle undergoes movement to ForceP in narrow syntax. Nevertheless, the particle appears in clause-internal position on the surface, on the grounds that the initial, but not the topmost, copy of the particle is pronounced. (In (43), the copy indicated by strikethrough is not pronounced, although it occurs there in narrow syntax.)

Now, the absence of a wh-island effect in (38b) suggests that in narrow syntax, the clause-internal focus particle də does not appear in ForceP of the wh-complement clause. If də is not located in ForceP in narrow syntax, long distance pseudo-clefting is possible, as illustrated by the derivation in (44a).

In (44a), the null pseudo-cleft operator can be moved to its scope position via ForceP in the embedded clause. Furthermore, in (44a), the interrogative particle də, which is not located in its scope position, needs to undergo LF movement to ForceP. The intermediate copy of the null operator in ForceP created by pseudo-clefting in (44a) is not necessary for interpretive purposes and thus it is deleted at LF (cf. Lasnik & Saito 1992; Chomsky 1995). Once the intermediate copy is deleted, ForceP is available for LF focus particle movement. Since the derivations in (44) are both legitimate, (38b) is acceptable.

The data in (38) suggest that the interrogative focus particle də appearing in a wh-complement clause selected by the verb dannəwa ‘know’ may be moved to ForceP either in narrow syntax or at LF. If an operator feature in the Force-head triggers movement of a focus particle, it can be stated that in (38a), the operator feature to induce overt syntactic movement is assigned, but that in (38b), the operator feature is of the type that induces movement at the LF level. Chomsky (1995) postulates a strong versus weak distinction of formal/grammatical features; a strong feature needs to be eliminated in narrow syntax to avoid a crash at the PF level. On the other hand, such a restriction is not imposed on a weak feature, and its elimination can be delayed until LF. In (38a), overt syntactic movement is induced due to a strong operator feature assigned to the Force head, but in (38b), LF movement takes place by virtue of a weak operator feature assigned to the Force head.

Kishimoto (2005) suggests that in wh-interrogative focus concord constructions, the difference in the overt position of the interrogative focus particle də is conditioned by a semantic implication associated with the clause. For instance, in the direct wh-question in (45), where də appears at the end of clause, the speaker asks the hearer to provide the value of the wh-phrase, on the presumption that there is an individual that came.

Likewise, the wh-complement clause in (46a) carries the implication that there is an individual who came.

By contrast, the wh-complement clause in (46b) carries no such implication; the statement expressed in the embedded clause is neutral as to whether there is a person who came. In this case, the set of individuals that came could be empty, i.e. the proposition that someone came may be true or false.¹³

If the overt constituent position of the particle də in wh-interrogative focus constructions is conditioned by the semantic implication mentioned above, it is easy to see that in the type of wh-interrogative clause with the existential implication, the Force head is assigned a weak operator feature that allows its operator to be moved at the LF level.¹⁴ In this variant of the focus concord construction, a focus particle does not appear in ForceP in narrow syntax, so long distance A’-movement of DP via pseudo-clefting does not incur a wh-island violation, as in (38b). On the other hand, if this semantic implication is lacking, the Force head is assigned a strong operator feature that motivates the overt syntactic movement of an interrogative particle. In this case, the focus particle is moved to ForceP in narrow syntax, and hence, long distance A’-movement gives rise to a wh-island effect, as shown in (38a).

In the yes-no interrogative clauses in (47)(=(22)), wh-island effects are caused regardless of whether the particle də is positioned in clause-internal position or at the clause end.

The examples in (47) with yes-no interrogative complements differ from those involving wh-complements in (38), in that a wh-island effect obtains regardless of the position of the interrogative focus particle.

When the particle is placed clause-finally, as in (47a), a focused constituent is not syntactically delimited, as discussed in section 2. Thus, I hypothesize that də is directly inserted into ForceP to serve as an interrogative operator. If the clause-final focus particle də, which specifies the clause as interrogative, is positioned at the end of the clause, i.e. in ForceP, the derivation in (48) can be posited for (47a).

In (48), the movement of the null operator induces a wh-island violation, since the operator cannot stop at ForceP on its way, which is filled by də. Thus, a wh-island effect is observed in (47a).

Similarly, a wh-island effect is observed in (48b), where the focus particle də appears in clause-internal position. Note that an island violation is incurred when də is embedded in a syntactic island, as in (49)(=(33a)).

The fact suggests that in the yes-no interrogative clause in (47b), the focus operator də occupies ForceP, as a result of overt syntactic movement, even if the particle is phonetically realized in the base position. In (47b), null operator movement for pseudo-clefting induces a wh-island effect, owing to the focus marker residing in ForceP, as (51) illustrates.

The facts point to the conclusion that in (47b), where də appears clause-internally, the focus particle undergoes overt syntactic movement, while the initial rather than the final copy of the particle is pronounced.

The wh-island effects observed in (47) suggest that in the yes-no interrogative focus concord constructions in (47a) and (47b), də appears in ForceP in narrow syntax. This suggests that ForceP contains a strong operator feature to attract the interrogative particle də in narrow syntax regardless of whether də appears clause-internally or clause-finally on the surface. This being the case, the structure in (51a) can be posited for the complement clause in (47a), and (51b), for the complement clause in (47b).

Specifically, in (47b), the particle də, which delimits a focused constituent, is raised to ForceP in narrow syntax to eliminate a strong operator feature, but still, the particle is necessary for signaling what constituent is to be focused, i.e. to divide the presupposition and focused constituents. Since the particle in base position—the copy left in the initial position by particle movement—serves to fulfill this function, the tail member of the chain is pronounced, as in (51b), even though this type of phonetic realization is a marked option.¹⁵ The interrogative particle də occupies ForceP in (51b), and thus, in (47b), a wh-island effect is observed for long distance A’-extraction from the clause, although də appears clause-internally in the surface strings.

In declarative/non-interrogative focus concord constructions, no wh-island effect is obtained when long distance pseudo-cleft movement is invoked. The result is the same whether the focus particle tamay is placed in clause-internal or in clause-final position, as shown in (52).

The absence of a wh-island effect comes from the fact that the focus particle fills the lower operator position of FocP and does not move up to ForceP. Since clause-final tamay, just like clause-final də in yes-no interrogative constructions, does not delimit a focused constituent syntactically, it can be hypothesized that in (52a), tamay is base-generated in the scope position of FocP, and thus, the derivation in (53) can be posited for (52a).

The focus particle does not have an interrogative import, so that it does not move to ForceP. In consequence, tamay does not block the long distance movement of the null operator, which moves through ForceP on its way, i.e. the null pseudo-cleft operator can move to its scope position by way of ForceP, without violating the wh-island constraint.

When tamay appears clause-internally, focus is confined to the constituent to which it attaches. In this case, tamay is generated in clause-internal position and delimits a focus constituent syntactically. The clause-internal tamay needs to move to its scope position. In the present perspective, there are two possible derivations: one derivation involving overt constituent movement, and the other involving LF movement. Nevertheless, given that declarative focus constructions generally pattern with yes-no interrogative focus constructions, it is expected that tamay will move in narrow syntax even if it is pronounced in clause-internal position. If tamay is moved in narrow syntax, the derivation in (54) can be posited for the sentence in (52b).

In (54), the declarative particle tamay appears in FocP as a consequence of overt syntactic movement. In (54), ForceP, which serves as an intermediate landing site of the null operator moved by long distance pseudo-clefting, is not filled by tamay. If a clause-final declarative focus particle fills FocP, located lower than ForceP, the declarative particle and the null operator do not compete for the same position. The null operator can move through ForceP. Accordingly, the derivation in (54) is legitimate. In declarative/non-interrogative focus concord constructions, no wh-island violation is incurred by long distance pseudo-cleft extraction regardless of whether the focus particle appears clause-internally or clause-finally in the embedded clause.

There is good reason to believe that A’-movement of tamay takes place in narrow syntax rather than at LF. Evidence in support of this analysis may be adduced from the contrast in acceptability between (55a) and (55b).

In both sentences in (55), the wh-phrase takes embedded scope, i.e. the embedded clause is construed as an indirect wh-question. (55a) is acceptable, and tamay can take matrix scope. By contrast, (55b) is unacceptable, and tamay is prevented from taking matrix scope. The difference in acceptability emerges depending on the position of the interrogative particle də.

The data in (55) suggest that tamay undergoes movment in narrow syntax, but that the topmost copy of tamay is not pronounced. If tamay, just like a pseudo-cleft operator, undergoes long distance movement, it needs to move through ForceP in the embedded clause, as ForceP offers an escape hatch for long distance operator movement. The interrogative particle də in (55b) appears in ForceP, but in (55a), it appears in clause-internal position (in narrow syntax). When də appears in clause-internal position, as in (55a), tamay can undergo overt syntactic long distance extraction to the matrix FocP via the embedded ForceP without causing a wh-island violation, as illustrated in (56a).

Since the copy of tamay in the embedded ForceP is not needed for LF interpretation, it can be deleted, and də can move into that position at LF, as illustrated in (56b). Thus, (55a) is legitimate on the relevant interpretation. In (55b), by contrast, the long distance movement of tamay in narrow syntax is blocked by the particle də appearing in ForceP in the embedded clause, as illustrated in (57).

Consequently, (55b) displays a wh-island effect. In essence, the contrast in acceptability between (55a) and (55b), which emerges according to whether or not də appears in ForceP in narrow syntax, shows that the non-interrogative focus particle tamay undergoes overt syntactic movement even if it is pronounced in clause-internal position.

The present proposal gains further support from the fact that tamay cannot take matrix scope when the embedded clause is a yes-no question.

The unacceptability of (58) accrues from the fact that the interrogative particle də is overtly moved to ForceP in a yes-no interrogative clause even when it is realized in clause-internal position. In (58), just like (55b), the long distance movement of tamay gives rise to a wh-island effect, obtaining a configuration like (59).¹⁶

In light of the acceptability of (55a), coupled with the unacceptability of (55b) and (58), it can be concluded that the non-interrogative particle tamay undergoes overt syntactic movement even when it is realized in clause-internal position.

Finally, let us discuss the question of how focus falls on any constituent located within IP in yes-no interrogative focus constructions and in non-interrogative focus concord constructions, when their focus particles occur in clause-final position (see section 2). In both types of focus concord constructions, the clause-final particles encode scope, but do not delimit focused constituents. I surmise that in such cases, their focused constituents are fixed by way of “association with focus” (see e.g. Jackendoff 1972). As often discussed (e.g. Rooth 1985; Krifka 2006), it is generally the case that a focus operator can be semantically associated with a (non-adjacent) focused constituent as long as the latter falls under the scope of the former. In yes-no interrogatives and non-interrogative focus concord constructions, clause-final focus particles are focus operators located in the CP domain, so that they can be associated with any constituent below IP. Accordingly, when the focus particles appear in clause-finally, they can invoke a number of different interpretations. In the present perspective, the clause-final particles in yes-no interrogative and non-interrogative focus concord constructions are generated directly in their scope position, and mark only their scope, so that their focused constituents are determined in semantic/pragmatic terms.

The data on long distance pseudo-cleft extraction suggest that focus particle movement may be invoked either in narrow syntax or at LF in cases where the interrogative focus particle də appears in clause-internal position on the surface. In wh-interrogative clauses, the clause-internal focus particle də undergoes LF movement. In yes-no interrogative clauses, the interrogative particle də fills ForceP, even when it is not pronounced in that position. A clause-internal non-interrogative particle like tamay, just like the clause-internal də in yes-no interrogative clauses, undergoes overt syntactic movement to the CP domain. Nevertheless, in non-interrogative focus concord constructions, wh-island effects are not exhibited for long distance pseudo-cleft extraction, because the focus particle fills FocP, which is projected below ForceP (although it undergoes overt syntactic movement to the CP domain).

In a nutshell, focus particles in focus concord constructions display behavioral differences, depending on (A) whether focus particles are moved to their scope position in narrow syntax or at LF, and (B) where focus particles are located in the CP-domain. Specifically, if focus particles occupy ForceP in narrow syntax, they block long distance A’-movement out of the clause, but when they occupy FocP, no wh-island effects are observed for long distance A’-movement.

3.4 Implications

In Sinhala focus concord constructions, when a focus particle is placed inside syntactic islands, island effects are observed regardless of the type of particle, which suggests that focus particle movement is responsible for the island violations. Furthermore, the facts of long distance pseudo-cleft extraction in focus concord constructions suggest that focus particles show a difference in the timing of their particle movement, or that focus particle movement may take place either in narrow syntax or in LF. In wh-interrogative focus concord constructions, the interrogative particle də undergoes LF movement, i.e. the interrogative focus particle appearing in clause-internal position does not occupy the scope position in narrow syntax. In yes-no interrogative focus concord constructions, by contrast, ForceP is filled by the interrogative particle də even when it is superficially placed in clause-internal position. When the focus particle də appears clause-internally in yes-no interrogative constructions, its initial copy left by movement is pronounced.

In Sinhala focus concord constructions, both overt syntactic movement and LF movement are available for specifying scope for a focused constituent. In particular, in wh-interrogative focus concord constructions in (31), the clause-internal interrogative focus particle də moves to its scope position at LF, and thus, LF movement is responsible for the island violations. In yes-no interrogative focus constructions in (33), by contrast, the clause-internal focus particle də specifying the interrogative force of the clause undergoes movement in narrow syntax, which suggests that the island violations must be due to overt syntactic movement. The data suggest that both LF movement and overt syntactic movement give rise to exactly the same island violations, or that both types of movement are constrained by the island constraints in the same way.

Sinhala is a CP recursion language where ComplP, accommodating an ordinary complementizer, is ordered above the projections hosting the focus particles, i.e. ForceP and FocP. The wh-island effects observed for long distance A’-movement (pseudo-cleft extraction) suggest that ForceP serves as an intermediate landing site, i.e. long distance A’-movement needs to go through ForceP, which can accommodate a focus particle specifying the interrogative force of the clause. The wh-island effect does not obtain in the non-interrogative focus concord constructions where the focus marker occupies FocP, located in a position lower than ForceP filled by the interrogative focus marker də. The discrepancy in the wh-island effect observed between the interrogative and non-interrogative focus particles suggests that the possibility of long distance A’-movement is determined according to whether it can move through ForceP. This fact in turn suggests that the Minimal Link Condition (Chomsky 1995) or the Minimality Condition (Rizzi 1990), which is defined in terms of c-command without referring to a particular syntactic position, does not provide a full account for the facts of the island effects observed in Sinhala long distance pseudo-cleft extraction.

In the literature, there are attempts to reanalyze LF movement as an instance of overt syntactic movement (Brody 1995; Groat and O’Neil 1996; Pesetsky 1998; Bobaljik 2002). These attempts are largely motivated by a recent trend in generative literature—the phase-based theory, which postulates that the domain transferred for spell-out is inaccessible to syntactic operations or computations (Chomsky 2001; 2004; 2008), Nevertheless, the facts of Sinhala focus concord constructions illustrate that focus particle movement can take place both in narrow syntax and at LF. In particular, given that LF movement and overt syntactic movement bring out different syntactic consequences with regard to long distance pseudo-cleft extraction in focus concord constructions, the fact suggests that the two movement options cannot be reduced to one.