2.1. Two possible syntaxes behind preverbal foci

A language that is canonically used to illustrate the phenomenon of immediately preverbal focus is Hungarian (Uralic). In Hungarian, immediately preverbal placement of narrow foci is obligatory; example (3) below provides a construction with immediately preverbal object focus and demonstrates that inserting any other material – such as the adverb végül ‘eventually’ – between the focus and the verb leads to ungrammaticality.

Another language with obligatorily immediately preverbal focus is Georgian (Kartvelian), as illustrated for subject focus in (4):

Despite surface similarities between languages with preverbal focus placement, two distinct types of syntax have been identified as deriving the left-adjacency of the focus to the verb in these cases. Hungarian and Georgian, in fact, represent these two mechanisms. The first mechanism is (i) movement of both the focus and the verb into the specifier and head of the same functional projection (Bródy Reference Bródy1990a on Hungarian; Bhatt Reference Bhatt1999 on Kashmiri; Jayaseelan Reference Jayaseelan2001 on Malayalam; Elordieta Reference Elordieta2001 on certain types of foci in Basque; a.o.). As motivation for movement, feature-checking in Spec,XP or alignment of the focused constituent with nuclear stress/nuclear pitch accent (NPA), which coincides with the Spec,XP position, have been invoked; these correspond to Bródy’s (Reference Bródy1990a) and Szendrői’s (Reference Szendrői2003) analysis of Hungarian, respectively. The second mechanism of achieving focus–verb adjacency is (ii) displacement of the material intervening between the focus and the verb, via (post-)syntactic movement or base generation of a non-canonical word order. The displacement may be motivated by the information-structural properties of the displaced material (Şener Reference Şener2010 on Turkish; Borise Reference Borise2019; Reference Borise2023 on Georgian) and/or by the need for the focused constituent to carry the NPA (Arregi Reference Arregi2002 on Basque).

Whether a language with preverbal focus uses mechanism (i) or (ii) may be revealed, albeit in a language-specific manner, by a number of properties associated with the construction. One of these is verb inversion, which is symptomatic of preverbal focus of type (i) but not of type (ii). Verb-inversion phenomena encompass all cases in which an element that is rigidly preverbal in a canonical order appears postverbally in non-canonical orders and in which rightward movement of that element is ruled out. The best-known cases include the inversion of the (auxiliary) verb across the (clitic) subject in languages like English and French. Not all verb inversion operations involve inversion with a syntactic phrase, however. EA displays inversion of the auxiliary with the lexical verb, for instance (see Section 3.2 below). In Hungarian, too, one of the facts supporting a type (i) analysis of preverbal focus is verb inversion (Bródy Reference Bródy and Kenesei1990b). In the canonical order of broad focus utterances like (5a), so-called verbal modifiers, notably including verbal particles (prts), occupy an immediately preverbal position. By contrast, in sentences with a preverbal narrow focus the verb surfaces to the left of the verbal modifier, as in (5b).

Another clear diagnostic is long movement of the focused phrase. If a focused constituent originating within an embedded clause is displaced by a movement operation to the immediate left of the matrix verb, then this provides evidence of a raised preverbal focus of type (i) over the in situ focus construction of type (ii). The fact that Hungarian permits long focus movement (É. Kiss Reference Kiss and Katalin2002: 250–258 ), as illustrated in (6), provides justification to assigning its preverbal focus to type (i). EA sides with Hungarian in this regard too (see Section 3.2).

While surface verb inversion and long movement are striking symptoms of a structure of type (i), they are not necessarily found in every language with this kind of preverbal focus construction (e.g. verb movement may happen to be string-vacuous, and focus raising may happen to be clause-bound in a language). Consequently, their absence does not, strictly speaking, preclude a type (i) analysis.

What diagnostics may be both applicable and revealing in a language is a matter for language-specific empirical investigation. For instance, as syntactic movement, focus fronting may be expected to display island effects (provided that the relevant non-clausal islands can be constructed) or create cross-over effects. If such expected signs of focus movement are systematically lacking, then, unless their absence can be accounted for on the basis of other, independent properties of the language (such as the application of optional scrambling that obviates the crossing configuration that would give rise to a cross-over effect), this points to an in situ focus construction of type (ii).

This can be illustrated with Georgian, in which neither the focus nor the verb undergoes dedicated focus-related movement – instead, the focus–verb adjacency results from the displacement of intervening material to the left/right periphery. The cumulative evidence for this comes from, among others, the lack of movement effects associated with the focused constituent (e.g. island effects), the strict ban on cross-clausal focus movement, as well as the relative positioning of narrow foci and non-topicalized low adverbs (Borise Reference Borise2019, Reference Borise2023). For the sake of illustration, the latter is shown in (7): the focused constituent being placed below iʃviatad ‘seldom’ points to it being found low in the clausal structure – i.e. lower than the clause where a discourse projection FP may be found:Footnote ⁴

Georgian also offers an additional, language-specific test for the position of focus, which is based on the relative positioning of foci and negative indefinites. Specifically, it can be shown that negative indefinites in Georgian stay in their base positions and do not undergo A- or A-bar movement. When co-occurring with narrow foci, the relative order of negative indefinites and narrow foci is only compatible with in situ placement of narrow foci as well – as opposed to movement of foci to FP; for details, see Borise (Reference Borise2019, Reference Borise2023).

Taken together, these facts demonstrate that Georgian foci remain in situ, and the adjacency with the verb is achieved via displacement of the would-be intervening material. This is illustrated in (8):

To recap, despite the surface similarity, preverbal foci may be derived via two distinct syntactic configurations. It might seem that the two types of preverbal foci represent coincidentally identical outcomes of two unrelated grammatical processes, but we propose that a unified account rooted in the prosodic requirements of foci is possible.

2.2. The interface between information structure and prosody: The Focus-as-Alignment model

The analysis that we propose is built on bringing together two existing proposals. The first one is the FA model, proposed in Féry (Reference Féry2013). The main insight of the FA model is that focused constituents are required to be aligned with the right or left edge of a prosodic domain – most commonly, an ι. In order to achieve this alignment, a language may employ a number of strategies, including syntactic movement (and, possibly, post-syntactic displacement), an alternative syntactic structure (e.g. a cleft), insertion of a prosodic boundary, or enhancement of an existing prosodic boundary. According to the FA model, prosodic prominence, which is often taken to be a universal prosodic property of focus (Truckenbrodt Reference Truckenbrodt2005; Büring Reference Büring, Zimmermann and Féry2010; a.o.) may, but does not have to, co-occur with prosodic alignment. In other words, prominence and alignment of foci with prosodic boundaries are taken to be independent from each other (it is worth noting, though, that prosodic prominence also preferentially aligns with prosodic boundaries; more on this below).

The FA approach goes against the traditional assumption that prominence, in the form of the NPA, is an intrinsic, and/or the sole, required prosodic correlate of focus; this is the key tenet of the Focus-as-Prominence (FP) model (Jackendoff Reference Jackendoff1972; Reinhart Reference Reinhart1995; Truckenbrodt Reference Truckenbrodt1995; Zubizarreta Reference Zubizarreta1998; Gussenhoven Reference Gussenhoven2008; Büring Reference Büring, Zimmermann and Féry2010; a.o.). At least two pieces of evidence provide support for the FA as opposed to the FP approach. The first one is the existence of languages in which foci are aligned with prosodic boundaries but do not carry the NPA, because it is realized elsewhere in the clause: e.g. Nɬeʔkepmxcin/Thompson River Salish (Koch Reference Koch2008a, Reference Koch, Chang and Haynie2008b); see also Hoot (Reference Hoot, Geeslin and Díaz-Campos2012) for stress–focus mismatch in the case of focus on prenominal modifiers in Spanish. The second one is languages that do not provide evidence for a phonologically meaningful NPA, such as French (Féry, Hörnig & Pahaut Reference Féry, Hörnig, Pahaut, Gabriel and Lleó2010) and Georgian (Dzidziguri Reference Dzidziguri1954; Alkhazishvili Reference Alkhazishvili1959; Zhghenti Reference Zhghenti1963, Reference Zhghenti1965a).

Féry (Reference Féry2013) formalizes the FA approach as a series of violable constraints within the framework of OT. The key constraints are two Align constraints (McCarthy & Prince Reference McCarthy, Prince, Booij and van Marle1993; Selkirk Reference Selkirk1996; a.o.) that are responsible for the alignment of the focused constituent with an ι-boundary, as in (9). The higher-ranked one of the two determines whether a given language displays right- (9a) or left-alignment of focus (9b).

The FA approach also incorporates the view that alignment of a prosodic boundary with the main locus of prosodic prominence (e.g. an NPA) is enforced by a set of dedicated constraints (Truckenbrodt Reference Truckenbrodt1995; Samek-Lodovici Reference Samek-Lodovici2005). Here we adopt the Align-head constrains in (10), where head refers to the NPA.Footnote ⁶ As with the Align-focus constrains, higher ranking of one of these constraints ensures that it is dominant in a language.

Crucially, the lack of a direct link between focus and prominence (i.e. the absence of constraints directly aligning the narrowly focused constituent with nuclear prominence) in the FA approach means that the alignment of focus with prominence is mediated by alignment of both with a prosodic boundary – i.e. it is obtained in those languages where both (9a) and (10a), or (9b) and (10b), are high-ranked. Accordingly, the FA approach can successfully account for languages like Nɬeʔkepmxcin/Thompson River Salish, in which foci and prominence align with different edges of ι (left and right, respectively; Koch Reference Koch2008a, Reference Koch, Chang and Haynie2008b), and languages like French and Georgian, in which foci are not aligned with nuclear prominence due to their lack of a phonological notion of a NPA.Footnote ⁷

In addition to the Align constraints, Féry (Reference Féry2013) employs a lower-ranked constraint that penalizes word order permutations that deviate from the canonical word order (CWO) in a language, as shown in (11). Lower ranking of CWO means that it is violated in languages that employ specific syntactic configurations for the expression of narrow focus.

To illustrate the workings of the FA approach, consider the example of Italian, in which narrow foci (e.g. those found in responses to wh-questions) surface clause-finally and carry prosodic prominence (Samek-Lodovici Reference Samek-Lodovici2005; cf. also Szendrői Reference Szendrői2001), as shown in (12a). Crucially, placing the narrowly focused constituent anywhere other than the clause-final position – even if that corresponds to a canonical subject-verb (SV) word order, as in (12b) – results in infelicity.

Féry (Reference Féry2013) analyzes Italian as a language with right-aligned foci and right-aligned prominence, which translates into high-ranked Align-R(focus, ι) and Align-R(head, ι), which are unranked with each other but both outrank CWO. This is illustrated in (13); here and in other tableaux, focus is indicated by small caps and the NPA by boldfacing. In (13), candidate (a) satisfies both Align-R(focus, ι) and Align-R(head, ι), and the violation on CWO that it incurs is not fatal. Candidate (b) violates both higher-ranked constraints, and candidate (c) violates Align-R(focus, ι); these violations of the higher-ranked constraints take both (b) and (c) out of the race.

In contrast with Italian, Hungarian is known to align foci with the left edge of an ι, while also making them prosodically prominent via alignment with an NPA (Szendrői Reference Szendrői2001, Reference Szendrői2003), as in (14a). Note that, as in (5) above, the verb undergoes inversion over the verbal particle – here, fel. Deviating from this pattern – including the unmarked, broad focus word order (without the verb–particle inversion), as in (14b) – results in infelicity in this context.

According to the FA approach, Hungarian is a mirror image of Italian, in that both foci and prosodic prominence are subject to left alignment, due to high ranking of Align-L(focus, ι) and Align-L(head, ι). Both of these constraints, again, outrank CWO; note that Féry (Reference Féry2013) follows (Szendrői Reference Szendrői2003: 64) in assuming that the CWO within the VP/PredP in Hungarian is VSO.Footnote ⁸

To sum up, the FA approach derives focus placement and focus prominence from alignment with prosodic boundaries and successfully accounts for focus placement in a variety of languages. That said, the FA model alone cannot account for immediately preverbal focus placement (nor was it intended to). It offers no principled account of syntax–prosody mapping between ι and syntactic constituents (e.g. in Hungarian ι may correspond to PredP or FP) or focus–verb adjacency as such.

2.3. The syntax–prosody interface: The flexible ɩ-mapping hypothesis

ιs are commonly taken to correspond to syntactic ‘clauses’, but there is no unanimity on the exact syntactic counterpart of ι: e.g. CPs (Truckenbrodt Reference Truckenbrodt2005), TPs (Zerbian Reference Zerbian2006), and phases (Cheng & Downing Reference Cheng and Downing2007) have been proposed to systematically correspond to ιs. In this paper, we adopt the flexible ɩ-mapping hypothesis (Hamlaoui & Szendrői Reference Hamlaoui and Szendrői2015), which provides a general account of the cross-linguistic variation in the size of the ι containing the core clause. According to it, the size of the ι – the ‘core’ ι for our purposes – is determined syntactically: ι corresponds to the highest projection that hosts overt verbal material – ‘the verb itself, the inflection, an auxiliary, or a question particle’ (Hamlaoui & Szendrői Reference Hamlaoui and Szendrői2015: 80) (henceforth: HVP, for highest verbal projection). This is schematized in (16a), where the bracketed part of the tree, corresponding to the HVP, is labelled ι. This means that the size of the core ι of a clause is flexible and does not rigidly correspond to any predetermined syntactic projection.

Elements that are syntactically outside the HVP (including material projected above, or adjoined to, the HVP) are mapped to prosodic positions outside the core ι. In main clauses, these elements are included in a more encompassing, maximal ι (16b), since Hamlaoui & Szendrői (Reference Hamlaoui and Szendrői2015) adopt the hypothesis that constituents expressing an entire illocutionary act are also mapped to an ι (Selkirk Reference Selkirk, Frota, Vigário and Freitas2005). This leads to there being a nested ι-structure, with the maximal ι undominated and embedding the core ι (see Selkirk Reference Selkirk, Goldsmith, Riggle and Alan2011; Ito & Mester Reference Ito, Mester, Borowsky, Kawahara, Shinya and Sugahara2012, Reference Ito and Mester2013 for similar proposals).

The alignment of the HVP with ι is enforced by the two syntax–prosody mapping constraints in (17); the corresponding prosody–syntax mapping constraints are low-ranked and not listed here.

To illustrate, in Hungarian, narrow foci undergo movement to a functional projection FP, accompanied by raising of the verb to the head of the same projection, as discussed in more detail in Section 2.1. This means that when FP is projected, by virtue of the verb raising to F⁰, FP corresponds to the HVP, and, as such, it gets mapped to an ι. Hamlaoui & Szendrői (Reference Hamlaoui and Szendrői2015) follow the Focus-to-Accent/ Focus-as-Prominence view in assuming the Stress–Focus Correspondence principle, which requires focus to contain the main ι-level prominence (for a classic formulation, see Reinhart Reference Reinhart1995). Taking this together with the assumption that in Hungarian the main ι-level prosodic prominence is left-aligned within ι (Szendrői Reference Szendrői2003), it is explained why narrow focus is raised to Spec,FP: because in this position it will contain the main prominence of ι. In contrast to narrow focus, left-peripheral topics are structurally higher than the HVP and, accordingly, outside of the core ɩ. Topic- and focus-placement are schematized in (18a). Similarly, in a broad-focus clause, where the HVP corresponds to PredP, ɩ is co-extensive with PredP, while the verbal arguments further to the left are outside of it. This is shown in (18b). The assumed intonational phrasing is supported by accentuation patterns in declaratives and in polar interrogatives (Hamlaoui & Szendrői Reference Hamlaoui and Szendrői2015).

Notably, the specifier of the HVP is part of the ι, but the proposal makes it clear that the specifier of the HVP does not have to be occupied by a focal constituent, since topical constituents are also allowed in the HVP – e.g. in Bàsàá (Bantu). The authors are explicit that ‘it is not the topical or focal nature of a peripheral element that primarily determines whether it is phrased inside or outside the core intonational phrase […] rather, phrasing is dependent on the highest overt position of the verb’ (Hamlaoui & Szendrői Reference Hamlaoui and Szendrői2015: 83).

3.1. Hungarian

Our account of Hungarian is similar to the one in (15) in that it involves the constraints Align-L(focus, ι) and Align-L(head, ι). Crucially, though, it also includes the constraint in (17a), which mandates that the height of the left ɩ-edge corresponds to the left edge of the HVP. Accordingly, instead of stipulating the correspondence of ɩ-size to different syntactic projections in different contexts (e.g. FP in narrow focus contexts), essentially following Szendrői’s (Reference Szendrői2003) and Hamlaoui & Szendrői’s (Reference Hamlaoui and Szendrői2015) insight, we can derive this correspondence in a principled way. Most importantly, preverbal focus placement in Hungarian is derived by virtue of the fact that focus is aligned with the left ɩ-edge, which, in turn, is determined by the structural height of the verb. The only syntactic position that satisfies this requirement is Spec,FP, which leads to immediately preverbal focus placement. This is shown in (21) for the example (20b); (20a) is provided for context.

Prosodically, the NPA on the leftmost constituent within the ι is represented by a HL* pitch accent (e.g. Szendrői Reference Szendrői2003), aligned with the initial syllable of the focused constituent, given that Hungarian has initial lexical stress. While the left edge of the core ι does not have its own marking (beyond the NPA aligned with it), prefocal material, if available, often ends in a high boundary tone (Surányi, Ishihara & Schubö Reference Surányi, Ishihara, Schubö, Alcibar and Prieto2012; Genzel, Ishihara & Surányi Reference Genzel, Ishihara and Surányi2015), which acts as an indirect marker of the upcoming left ι-boundary.

In our OT account, the Hungarian focus facts are derived via the interplay of Align-L(focus, ι), Align-L(head, ι) (where the head of ι is the NPA), and Align-L(HVP, ι). The constraints are unranked with respect to each other and not violated by the winning candidate, (21a).Footnote ¹¹ *FP is ranked below the other constraints, meaning that the violation incurred by (21a) is not consequential. In (21b, c), the verb has raised to PredP, with its specifier occupied by the prt. Note that we take violations of Align-L(focus, ι) and Align-L(head, ι) to be cumulative: in (21c), removing the narrowly focused constituent egy széket ‘a chair’ from the left ɩ-edge by one prosodic word, fel rúgott ‘kicked’, results in one violation of Align-L(focus, ι), whereas in (21b), removing it from the left ɩ-edge by two prosodic words, fel rúgott ‘kicked’ and a férfi ‘the man’, results in two violations of Align-L(focus, ι). This will be important in the account of in situ preverbal foci.Footnote ¹²

3.2. Eastern Armenian

EA, an Indo-European language spoken in the Caucasus, is another verb-final language that is described as having immediately preverbal focus (Comrie Reference Comrie1984; Dum-Tragut Reference Dum-Tragut2009; Megerdoomian & Ganjavi Reference Megerdoomian and Ganjavi2000). In a broad-focus EA clause, if the verbal complex consists of a participle and a copula, the copula follows the participle, as shown in (22); note that definite objects in EA typically surface postverbally (Megerdoomian Reference Megerdoomian2002; Dum-Tragut Reference Dum-Tragut2009: 562).

In the context of narrow focus, the copula raises past the participle and cliticizes to the narrowly focused constituent, as in (23). This inversion is obligatory and provides initial evidence in favor of EA having a raised preverbal focus configuration.

A definitive piece of evidence in favor of raised preverbal foci in EA comes from long focus movement. To illustrate, (24a) provides a broad-focus example of a clause with an infinitival complement. As (24b) shows, when the complement of the infinitive – here, t’atroni bemě ‘theater stage’ – is focused, it is raised out of the infinitival clause into the matrix clause and immediately followed by the inverted copula. We analyze this as movement of focus to a specifier of a high FP projection in the matrix clause, accompanied by movement of the copula to the head of the same projection.

A note about the status of the copula is in order. Traditionally, examples like (23) and (24b) have been treated as cases of mobile auxiliary cliticization, given the clitic-like behavior of the auxiliary (Tamrazian Reference Tamrazian1994; Kahnemuyipour & Megerdoomian Reference Kahnemuyipour and Megerdoomian2011). Most verbal tenses in EA, including simple present, are analytical and, thus, include the auxiliary. Nonetheless, the correct generalization is that preverbal foci in EA immediately precede finite verbs and do not necessarily involve auxiliaries (Pregla Reference Pregla2024). This can be demonstrated with synthetic tenses like aorist, shown in (25), in which the finite verb behaves in the same way as the auxiliary in (23) and (24b), immediately following the narrow focus. We take this to indicate that both auxiliaries and other finite verbs can occupy the head of the HVP in EA. Accordingly, we propose that in EA, similarly to Hungarian, foci are raised and left-aligned with an ɩ-edge, which corresponds to the HVP (namely, the FP projected to house focus).

Prosodically, narrow foci in EA carry an NPA of the H*+L shape, anchored to the stressed syllable, which is typically final or, if the final vowel is a schwa, penultimate in a prosodic word (Dum-Tragut Reference Dum-Tragut2009: 395, 619; Skopeteas Reference Skopeteas2021; Seyfarth et al. Reference Seyfarth, Dolatian, Guekguezian, Kelly and Toparlak2024). According to Chakmakjian, Dolatian & Skopeteas (Reference Chakmakjian, Dolatian and Skopeteas2024), the left edge of the prosodic constituent that contains the focus and the verb to the exclusion of other – e.g. prefocal – material (i.e. in our terms, the ι, though Chakmakjian et al. implicitly take it to be a φ) is marked by a low initial boundary tone -L and characterized by phrasing the narrowly focused constituent together with the verb (which is not the case in broad focus contexts).

The OT account is provided in (26). Because narrow foci in EA are prosodically prominent – i.e. they carry the NPA (Dum-Tragut Reference Dum-Tragut2009: 395, 619; Seyfarth et al. Reference Seyfarth, Dolatian, Guekguezian, Kelly and Toparlak2024) – the constraint Align-L(head, ι) is included in the tableau. As in Hungarian, the winning candidate honors the three Align constraints, unranked with each other, and violates the lower-ranked *FP. For the sake of brevity, we only consider candidates with the word order characteristic of preverbal narrow focus; modulo the relevant language-specific facts (i.e. inversion of the auxiliary with the lexical verb), the EA facts would pattern like the Hungarian ones in (21).

3.3. Iron Ossetic

Similarly to Hungarian and EA, IrO, a verb-final Iranian language spoken in the Caucasus, is described as placing foci immediately preverbally (Abaev Reference Abaev1939; Erschler Reference Erschler2008, Reference Erschler2012).Footnote ¹³ This is shown in (27b), with the broad focus counterpart provided in (27a) for comparison.

The evidence for the raised status of IrO preverbal foci comes not from the inversion of the finite verb with a particle or participle, like in Hungarian or EA, but from the interaction of narrow foci, wh-phrases, and negative indefinites. In IrO, each of these constituent types, if not co-occurring with others, must surface immediately preverbally. When co-occurring, though, the strict focus >> wh >> neg >>V order is obtained, regardless of e.g. thematic roles, as shown in (28).Footnote ¹⁴ This suggests that the elements in the preverbal cluster take up their surface positions via movement. Accordingly, we propose that immediately preverbal foci in IrO (even when not co-occurring with wh-phrases of negative indefinites) undergo movement to the specifier of a dedicated projection, accompanied by raising of the verb. Except for wh-phrases and negative indefinites, (practically) no other elements can intervene between the narrowly focused constituent and the verb.Footnote ¹⁵

On the prosody side, IrO preverbal foci are left-aligned with an ι-edge and carry an NPA H*, aligned with the initial or second syllable of the leftmost prosodic word of the focused constituent, depending on vowel quality. The left ι-edge itself is unmarked, but the topicalized constituent(s) that precede it, if any, carry a characteristic L+H* contour (Borise & Erschler Reference Borise and Erschler2022, Reference Borise and Erschler2023). The OT account of preverbal focus placement in IrO is provided in (30), deriving the example of preverbal subject focus in (29), repeated from (27b) for convenience.

To recap, for languages like Hungarian, EA, and IrO, it can be shown that immediately preverbal foci are raised and accompanied by verb movement, which yields focus–verb adjacency. We propose, building on Szendrői (Reference Szendrői2003), Féry (Reference Féry2013), and Hamlaoui & Szendrői (Reference Hamlaoui and Szendrői2015), that the movements in all these languages are motivated by the need for the focused constituent to left-align with the ι-edge, which is created by verb movement. Preverbal placement of raised foci, therefore, is the by-product of the need for adjacency with the ι-edge, which itself is derived with reference to the position of the verb.Footnote ¹⁶

4.1. Turkish

Along with Hungarian, Turkish is one of the most well-described languages with immediately preverbal focus placement (Erkü Reference Erkü1983; Erguvanlı Reference Erguvanlı1984; Göksel & Özsoy Reference Göksel, Özsoy, Göksel and Kerslake2000; İşsever Reference İşsever2003; Öztürk Reference Öztürk2004; Şener Reference Şener2010; Kamali Reference Kamali2011). Neutral broad-focus clauses in Turkish are verb-final, as in (31), while narrow foci are strongly preferred to be placed immediately preverbally, as in (31b).Footnote ¹⁹

In the context of narrow focus, the placement of prefocal – as well as postverbal – elements is determined by strict IS-requirements. Only contrastive and aboutness topics can precede preverbal foci, though, notably, the order of prefocal elements is flexible: cf. (31b) and (32):

When the preverbal element is focused, Turkish allows for constituents that represent given, backgrounded information to follow the verb. Their relative order is also flexible, as in (33). In the literature on Turkish, these facts – the IS-requirements on prefocal and postverbal material and the flexible relative order of constituents within the prefocal and postverbal field – have been taken to mean that the non-focal material undergoes movement to the peripheries of the clause; this leads to the adjacency between the focus and the verb, which are not subject to focus-related movement. There is agreement in the literature that Turkish does not have a dedicated focus projection (Butt & King Reference Butt and King1996; Göksel & Özsoy Reference Göksel, Özsoy, Göksel and Kerslake2000; Şener Reference Şener2010; a.o.).

In terms of general clausal architecture in Turkish, it has been argued that verbs raise to Asp⁰/T⁰, on the right of the clausal spine (Aygen Reference Aygen2004; Yarbay et al. Reference Yarbay, Tuba and Bastiaanse2008). Non-quantificational subjects raise to Spec,AspP, while Spec,TP is reserved for quantificational elements, which may or may not be subjects (Öztürk Reference Öztürk2004, Reference Öztürk2005). It has also been suggested that (non-clausal) postverbal constituents in Turkish result from movement to the right and land (and take scope) very high in clausal structure (Kural Reference Kural1997; Kornfilt Reference Kornfilt, Sabel and Saito2005; Şener Reference Şener2005; a.o.). If so, postverbal constituents are outside of the HVP, which corresponds to AspP/TP.

On the prosody side, based on the flexible ɩ-mapping hypothesis, this means that postverbal constituents are predicted to be outside of the core ɩ. That is, the core ɩ encompasses all clausal material other than the postverbal constituents (it also excludes left-peripheral topics, though this is less relevant for the present purposes). This correspondence between basic clausal syntax and prosody in Turkish is schematized in (34a) and illustrated in (34b):

Postverbal constituents in Turkish are strictly deaccented (Özge & Bozsahin Reference Özge and Bozsahin2010); even pitch accents on lexically accented words are removed (Güliz Güneş, personal communication, August 2023). There is strong evidence for an NPA, which targets the immediately preverbal position and is expressed as a H* pitch accent; the right edge of the core ɩ, aligned with the right edge of the verb, carries a low boundary tone LL% (even in the presence of postverbal material) (Özge Reference Özge2003).

We propose that in Turkish preverbal foci align with the right ɩ-edge.Footnote ²⁰ Our OT analysis, therefore, relies on the right-edge versions of the three key Align constraints employed in Section 3: Align-R(focus, ι), Align-R(head, ι), and Align-R(HVP, ι). Now that we are dealing with nested ɩs and several right ɩ-edge targets of alignment, as in (34), we need to fine-tune our analysis. To do so, we adopt the constraint Foc _ι-core (Borise Reference Borise2023), which is based on the insight in Szendrői (Reference Szendrői2003):Footnote ²¹

To model NPA placement, we employ a markedness constraint *X⁰-Stress, (36), inspired by Duanmu’s (Reference Duanmu2000: 146) descriptive rule Nonhead Stress. It captures the empirical generalisation that, in a configuration [_XP YP X⁰], the option for the syntactic head X⁰ to carry a (nuclear) pitch accent is marked, both typologically and language-internally – compared to the phrase YP doing so. Without (36), a right-aligning verb-final language would assign NPA to the verb.Footnote ²²

Our OT analysis is provided in (37). As before, the three Align constraints (this time targeting the right ɩ-edge) are unranked with respect to each other. Additionally, they are also outranked by *X⁰- Stress (*FP is left out here and below, as the relevant candidates do not differ in terms of the functional projections they contain). The winning candidate, (37a), incurs a violation of Align-R(focus, ι) but still fares better than its competitors. Candidates (37b) and (37d) lose because the violations of Align-R(focus, ι) and Align-R(head, ι) get worse with each prosodic word that separates the focus from the right ɩ-edge and with each pitch accent that separates the NPA from the right ɩ-edge. Candidate (37c) – or any other candidate with an accented verb – is suboptimal because it violates *X⁰- Stress, even though it aligns the NPA perfectly with the right ɩ-edge.

Our approach to Turkish can be extended to some other Turkic languages, though additional strategies would be needed to account for other language-specific focusing strategies – e.g. left-peripheral information focus in Uyghur (Çetinkaya Reference Çetinkaya2023).

4.2. Georgian

Georgian (Kartvelian) differs from the languages discussed here so far in that, even though the OV word order is understood to be the underlying one, neutral broad-focus declaratives can have an OV (object-verb) or VO (verb-object) word order (Aronson Reference Aronson1990; Harris Reference Harris, Sornicola, Poppe and Shisha-Halevy1993; Nash Reference Nash1995; Boeder Reference Boeder2005; Skopeteas, Féry & Asatiani Reference Skopeteas, Féry and Asatiani2009; Skopeteas & Fanselow Reference Skopeteas and Fanselow2010; a.o.), as in (40).

Like other languages discussed here, Georgian has a requirement for preverbal placement of narrow foci, as shown in (41):

In the context of preverbal narrow focus, like in Turkish, given material can also appear postverbally:

Georgian preverbal foci are similar to those in Turkish, in that they do not undergo any focus-specific movement, and the adjacency with the verb results from the intervening material moving out from between the focus and the verb into the left and right peripheries of the clause. This is based on several pieces of evidence, such as lack of island effects, lack of weak cross-over effects, and interaction with adverbs, all of which suggest that Georgian preverbal foci are found in situ; see Borise (Reference Borise2023, Reference Borise2019) for details. Another, Georgian-specific test for the in situ status of preverbal foci comes from the interaction between preverbal foci and negative indefinites (neg-words). Several pieces of evidence combine to show this. First, it is independently known that there is no movement for case in Georgian – i.e. case is assigned to nominals in situ (Legate Reference Legate2008; Nash Reference Nash, Coon, Massam and deMena Travis2017). Second, neg-words in Georgian have been shown to resist movement into the left or right peripheries of the clause (Borise Reference Borise2019, Reference Borise2023). Combined, these two facts suggest that neg-word subjects in Georgian are necessarily found in situ. As (43) shows, though, a narrowly focused object can only be found to the right of an in situ neg-word subject – i.e. quite low in the clausal structure. Since it would be unmotivated to postulate a discourse projection within the thematic domain of the clause, evidence of this sort also points to the in situ status of Georgian preverbal foci.

Georgian differs from Turkish in several important ways, though. As already mentioned, both OV and VO are unmarked in broad focus. Following previous work by Skopeteas & Fanselow (Reference Skopeteas and Fanselow2010), we take broad-focus, neutral VO to be derived via short movement of the verb over the object. This is in line with the approach in which neutral word orders can be derived via semantically vacuous movement of a syntactic head or a phrase that contains the head (Cinque Reference Cinque2005; Abels & Neeleman Reference Abels and Neeleman2012), as opposed to e.g. phrasal movement to the right (cf. Abels Reference Abels2016; Neeleman Reference Neeleman2017; Pregla Reference Pregla2024).

Georgian also has more than one option for narrow focus placement: In addition to immediately preverbal focus, it allows for posverbal foci. This is the case for both arguments and adjuncts. To illustrate, the main possibilities for object and subject focus placement are provided in (44) and (45), respectively. Note that, if the focus is postverbal, there is a preference for it to be the only postverbal constituent; if there are other postverbal constituents in the same clause (not all speakers allow for this), there is a preference for focus to be the right-most one.

Syntactically, postverbal foci have been shown to be derived via right-adjunction (Borise Reference Borise2019, Reference Borise2023).Footnote ²³ In contrast with elements moved to the right in Turkish, though, Georgian postverbal foci tend to take narrow scope with respect to left-peripheral material and TP-level adverbs, which suggests that they are not necessarily found very high in the clausal structure. Most importantly for our purposes, though, by virtue of being adjuncts, postverbal foci in Georgian are outside of the HVP, and, accordingly, outside of the core ɩ. This is schematized in (46). Note that in Georgian, unlike in Turkish, ‘postverbal elements’ in (46) can be represented either by given material, as was shown in (42), or narrow focus, as in (44c) or (45c).

Note also that we label the XP that corresponds to the core ɩ (i.e. the HVP) as ‘AspP/TP’ in (46). For several reasons, this choice is not obvious, but we believe that it is on the right track. It is not obvious because, as already mentioned, subjects do not raise to Spec,TP, and instead stay low in their base positions and are assigned case there (nominative, ergative, or dative; see Legate Reference Legate2008; Nash Reference Nash, Coon, Massam and deMena Travis2017). There is also contradictory evidence with respect to the height of the material that is topicalized/dislocated into the left periphery in Georgian (CP or below), which leaves the occupancy of Spec,AspP/TP unclear.

With respect to the position of the verb, we adopt the view that the verb in Georgian clause stays low, possibly as low as V⁰, and that the direction of the clausal spine switches from head-final to head-initial quite low – possibly at vP (Borise Reference Borise2019), which allows for short V⁰-to-v⁰ movement to derive neutral VO clauses, as discussed with respect to (40b). This, however, means that the right and left edges of HVP – and, accordingly, ɩ – are derived quite differently in Turkish and Georgian.

In Turkish, with the clause being head-final and the verb raising to Asp⁰/T⁰, the right edge of the HVP/ɩ corresponds to the raised position of the verb, while the left edge is determined by the left-hand specifier of the HVP; the HVP extends from Spec,HVP on the left to the head of the HVP on the right. This is schematized in (47a). In contrast, in Georgian, the right edge of the HVP/ɩ corresponds to the low position of the verb. Because the clause becomes head-initial at vP, the height of the left HVP-edge, however, is not immediately clear.

The fact that the verb stays low would speak in favor of a very small HVP – as small as VP. On the other hand, Georgian has famously well-developed aspectual and agreement verbal morphology, which must be generated in the higher functional heads (cf. Lomashvili Reference Lomashvili2011). To reconcile that with the low position of the verb, we assume that verbal morphology is combined with the verbal root via post-syntactic lowering (cf. Georgieva, Salzmann & Weisser Reference Georgieva, Salzmann and Weisser2021). However, we take the fact that e.g. aspectual/tense morphology is generated in Asp⁰/T⁰ to mean that the left edge of the HVP in Georgian corresponds to AspP/TP, given that the verbal morphology acts as the head of the highest projection that hosts overt verbal material (which is defined by Hamlaoui & Szendrői (Reference Hamlaoui and Szendrői2015: 80) as ‘the verb itself, the inflection, an auxiliary, or a question particle’; see Section 2.3 above). Accordingly, the prediction is that the HVP should not be smaller than AspP/TP. We assume, therefore, that the core ι in Georgian corresponds to the HVP as schematized in (47b).

We propose that Georgian foci are similar to Turkish in that they align with the right ι-edge.Footnote ²⁴ Georgian prosody, though, is markedly different from Turkish, since there is no evidence for a phonologically meaningful NPA (Dzidziguri Reference Dzidziguri1954; Alkhazishvili Reference Alkhazishvili1959; Zhghenti Reference Zhghenti1963, Reference Zhghenti1965b) – instead, the prosodic structure in Georgian consists of Accentual Phrases (APs) that, in broad focus contexts, are usually co-extensive with prosodic words, with each AP carrying a pitch accent and a boundary tone (Vicenik & Jun Reference Vicenik, Jun and Jun2014).Footnote ²⁵ In preverbal focus contexts, a narrowly focused constituent is phrased together with the verb into a single prosodic constituent, an Intermediate phrase, characterised by a high pitch accent H* on the narrowly focused constituent, a H+L phrase accent on the verb and a low final boundary tone, which co-occurs with and is overridden by L%, the low ι-final boundary tone (Vicenik & Jun Reference Vicenik, Jun and Jun2014). Postverbal focus contexts are characterized by a high final boundary tone H% on the core ι, a low pitch accent L* on the narrowly focused constituent, and a L% at the end of the maximal ι (Skopeteas et al. Reference Skopeteas, Féry and Asatiani2009; Skopeteas & Féry Reference Skopeteas and Féry2010). In OT terms, the absence of NPA means that *X⁰-Stress and Align-R(head, ι) are irrelevant to Georgian; for the sake of simplicity, we omit them in the tableaux.

The possibility of both preverbal and postverbal narrow foci, we argue, stems from the constraint Foc _ι‑core being unranked with respect to the remaining Align constraints, Align-R(focus, ι) and Align-R(HVP, ι) – instead of being ranked above them, like in Turkish. This approach allows us to derive both the preverbal and postverbal focus configurations, as illustrated e.g. in (44) and (45) for object and subject foci. This is shown in the tableaux (48) and (49), respectively.

Before wrapping up, we would like to point out an interesting prediction that our account makes. Because focus–verb adjacency in the case of in situ preverbal foci is achieved via displacement of intervening material, we predict that the elements that cannot move out (e.g. for independent syntactic reasons) would remain as interveners. We have some evidence that this is indeed the case. Recall that, in Georgian, negative indefinites are always found in situ. The prediction is that a preverbal subject focus will necessarily be separated from the verb by the negative indefinite object. This is indeed the case, as shown in (50a) (though subject to some interspeaker variation: Some speakers only allow the O_negVS_F alternative, with postverbal subject focus, which does not run into this problem). For comparison, an example with the reversed theta-roles, in which neither argument is in situ, as in (50b), is starkly ungrammatical.

To recap, we have shown that languages with in situ preverbal foci, like Turkish and Georgian, can be successfully modelled as languages in which foci are aligned with the right ɩ-edge. Here, the right-alignment is necessarily imperfect, in that the verb separates the narrowly focused constituent from the right ɩ-edge; still, other candidates fare even worse. Importantly for the overall topic of this paper, the obligatory adjacency of in situ preverbal foci and the verb proves to be circumstantial – i.e. there is nothing about the preverbal positioning per se that is beneficial for focus placement. The same can be said about the obligatory adjacency of raised preverbal foci and verbs – here, too, the adjacency results from independent syntactic and prosodic properties (i.e. verb raising determining the left ɩ-edge and the focus aligning with it).