Scholars acknowledge the existence of at least two main types of wh-in situ of the answer-seeking type, namely one which constitutes the only possible strategy of question-formation, as illustrated in the Chinese examples in (1), and one which co-exists peacefully with total wh-fronting into the HLP, as in the French examples in (2):

Canonical instances of languages of the first type are Asian languages such as Chinese, Korean, and Japanese, while the second type, which is commonly referred to as ‘optional wh-in situ’ is typically found in the Romance languages, although not exclusively.

However, there exists at least another type of ‘wh-in situ’, in which the clause-internal wh-element must move to a low internal-merge site. A movement of this type is observed in Eastern Trevisan (henceforth, ‘Trevisan’ for short), a Northern Italian dialect spoken in the Veneto region. An example is provided in (3):1

In Trevisan, apparent wh-in situ co-exists with total wh-fronting into the HLP, as in (4):

In the Romance literature, a low focalisation would typically be analysed as unmoved, with everything that follows it linearly either non-clitic right-dislocated or ‘marginalised’, in the sense of Cardinaletti (2001) and Samek-Lodovici (2015). However, I claim that this cannot be correct for orderings like (3), which are better analysed as a by-product of the movement of the wh-element, done under focus-agreement. This low movement is a cross-linguistically robust phenomenon, but to the best of my knowledge has never been proposed for Romance before. My claim that there can be low focus movement is thus not new per se, given that it has been made in several other contexts and supported by work on numerous languages (§2.3), but it is new for Romance, and has never been addressed within the framework that I adopt here.

The key contribution of this article is that it provides robust empirical evidence of low focus movement in Romance and analyses the phenomenon of Northern Italian wh-in situ in an innovative and cross-linguistically valid way, namely by implementing Cable’s (2010) work on the syntax of Q-particles in answer-seeking wh-interrogatives. This results in a framework for the study of wh-in situ that is not language- or family-specific, in contrast to what has been done in the Northern Italian/Romance literature so far. With minor modifications, my adaptation of Cable’s theory can explain the morphosyntax of clause-internal wh-elements in Indo-European, Indo-Aryan, Dravidian, Niger-Congo, Sino-Tibetan, Koreanic and Japonic varieties, at the very least.2

I will indeed build on the low movement of Trevisan clause-internal wh-elements to claim that the languages of the world can basically be split into two macro types: those that allow feature-bundles in the HLP, and those that scatter the features relevant to interrogative wh-movement between the LLP and the HLP. Accordingly, while some languages check both [foc(us)] and [q] in the HLP, other languages first check [foc] in the LLP, and then only [q] in the HLP.3 Whether the two types of languages display or not overt fronting of wh-elements is regulated by a movement parameter.

Following Cable (2010), I also assume that among these two major types two sub-types exist: that of languages in which the Q-particle projects, and that of languages in which Q adjoins to wh-elements in interrogatives. I then explain the optionality in the alternation between wh-in situ and wh-ex situ in languages of the ‘optional’ type in terms of the existence of various steps of a generalised linguistic evolution towards unmoved wh-in situ followed by overt movement of Q – a claim supported by the diachronic development of wh-in situ languages such as Chinese, Japanese and Sinhala (§4.2).

At first glance, to posit the existence of silent Q-particles in Trevisan might seem totally ad hoc. However, there are theoretically and empirically valid reasons why Q-particles ought to be cross-linguistically implemented in the derivations of answer-seeking wh-questions, including those of Romance.4 First, given that in the cartographic enterprise the existence of a functional head in one language is sufficient to posit the existence of that very same head in all languages, the attested availability of phonetically-realised Q-particles cannot be ignored. Second, Trevisan data on the interrogative strategy of subject-clitic inversion (SCLI) in the presence of long wh-extraction and of syntactic islands empirically support this theoretical implementation, as I claim in §3.2.1.

It must be noted that, derivations of the kind presented here are firmly anchored on the theoretical assumption that wh-elements are not inherently interrogative, and do not take part in clause-typing (Aboh & Pfau 2011). In this framework Cable’s (2010) proposal for an extension of interrogative Q-particles to all languages should be embraced not only for the sake of universality, but also because Q-particles are cross-linguistically needed for clause-typing.5 Accordingly, all previous analyses of (moved or unmoved) wh-in situ cannot be maintained, hence the importance of the theory that I develop here.

The paper is organised as follows. I first present and discuss novel data from Trevisan that support the existence of low movement of both wh-elements and contrastive foci in Romance, and then discuss cross-linguistic evidence that low focus movement is a robust phenomenon in many languages (§2). Subsequently, I provide an overview of the main ingredients of Cable’s (2010) work on Q-particles and propose an extension of it to Trevisan (§3).6 A close comparison between the morphosyntax of wh-movement and focus movement in Trevisan and standard Italian will then constitute the foundation of my theory of ‘interrogative feature scattering’ vs ‘feature bundling’ (§4). Finally, I propose an implementation of Cable’s interrogative parameters and outline an initial diachronic typology of wh-interrogatives in the languages of the world (§4).

Since Huang (1982), pure in situ languages like Chinese and Japanese have been argued to have real wh-in situ, i.e., wh-elements surface in their external-merge position and are generally incompatible with overt wh-fronting. Wh-elements are commonly understood as operators which bind variables at LF. Like other quantifiers, wh-operators must be split across two positions to be interpretable, one which serves as the operator and one as the variable. The implication of this understanding is that all wh-elements must move to create the relevant operator-variable configuration before interpretation. In wh-fronting languages like standard English, for instance, the configuration is obtained in overt syntax, as in (5):

Conversely, for Huang (1982) and related, in languages like Chinese the correct operator-variable configuration is obtained in covert syntax. The movement of the wh-element is done after Spell-Out, at the level of interpretation (LF). Accordingly, while overt wh-fronting is illicit in pure in situ languages, covert wh-fronting does take place, as in (6):

Huang’s influential proposal faces at least two sorts of problems. First, despite being subject to the same interpretative scope as their moved counterparts, clause-internal wh-elements are constrained differently than overtly fronted wh-elements in terms of sensitivity to islands and intervention effects. Despite various explanations within the Principles and Parameters approach, the difficulty of dealing with these challenges in the framework of a simple ‘LF vs Surface-structure’ phrasal-movement parameter led researchers to pursue alternative paths for capturing the differences between wh-in situ and overt wh-movement (refer to Cheng 2003 for a survey). A second problem posed by Huang’s generalisation is that there is an asymmetry between wh-movement languages such as English and wh-in situ languages like Chinese. Indeed, while the latter lack any vestiges of overt wh-movement, the former do not eschew the wh-in situ strategy. English cannot be said to manifest only the positive setting of the movement parameter, since wh-in situ is not only possible, but in limited cases even compulsory in the language, notably in multiple wh-questions. Therefore, more than a mere alternation between overt and covert wh-movement must be involved in wh-interrogatives.

In contrast to wh-in situ languages like modern Chinese, those Romance languages that allow wh-elements to surface clause-internally also display the option of total wh-fronting into the HLP, albeit to different extents. Although wh-fronting is always possible, wh-in situ can be limited to certain wh-elements, and its availability varies intra- and cross-linguistically. Optional in situ languages include French, NIDs, and partially also Spanish and Portuguese. The phenomenon of Romance wh-in situ is made intriguing by the morphosyntax of Trevisan, in which the alternation between wh-fronting and ‘wh-in situ’ features compulsory movement of clause-internal wh-elements, as in (3), repeated as (7):

Examples like (7) support my claim that wh-in situ in this variety is only apparent. Significantly, an equivalent instance of low movement can also be observed in the case of clause-internal contrastively focused constituents, as in (8):7

That the position targeted by the low movements in (7) and (8) lies at the edge of vP is straightforward. First, since vP is acknowledged to be a phase (Chomsky 2001), it is likely to have a periphery. Second, the Trevisan active past participle can be demonstrated to surface in a derived position outside vP (§2.2.2), whence the observed ‘past participle > wh-element’ order. Significantly, that low focus movement of wh-elements and/or contrastive foci is a robust phenomenon in many languages of the world (§2.3), will further argue against an analysis of wh-in situ as a mere instance of covert wh-movement. I shall therefore claim that the parallelism between the clause-internal movement of focused elements and that of wh-elements indicates that the latter are inherently focused and undergo low movement for focus purposes.

That the position targeted by the low movements in (7) and (8) lies at the edge of vP is straightforward. First, since vP is acknowledged to be a phase (Chomsky 2001), it is likely to have a periphery. Second, the Trevisan active past participle can be demonstrated to surface in a derived position outside vP (§2.2.2), whence the observed ‘past participle > wh-element’ order.

Significantly, that low focus movement of wh-elements and/or contrastive foci is a robust phenomenon in many languages of the world (§2.3), will further argue against an analysis of wh-in situ as a mere instance of covert wh-movement. I shall indeed claim that the parallelism between the clause-internal movement of focused elements and that of wh-elements indicates that the latter are inherently focused and undergo low movement for focus purposes. Therefore, while the data discussed in this paper do not make Trevisan exceptional or unexpected in the bigger cross-linguistic picture, the fact that a movement of this sort has never been empirically demonstrated in Romance makes it both typologically and theoretically relevant for the general understanding of wh-in situ.

My aims in the following sub-sections are twofold: first, to claim that the low movement under consideration is compulsory for wh-elements; second, to show that the targeted position is Foc, a criterial projection in the LLP.

That the position targeted by Trevisan clause-internal wh-elements is vP-peripheral, as discussed in §2.2.2, supports Manzini’s (2014) intuition that the low left-peripheral Foc is involved in the derivation of Northern Italian wh-in situ. Interestingly, in many languages what looks like wh-in situ has been claimed to be overt wh-movement to a low position to the left of vP. These languages have been described as a “typologically interesting and significant type between full moving and in-situ languages” (Cheng & Bayer 2017: 21), and authors such as Kahnemuyipour (2001) understand the low movements of wh-elements and foci as driven by a focus requirement. I survey some significant works in what follows; for a thorough review, see Bonan (2021: 91).

In Trevisan, contrastively-focused constituents surface clause-internally, mostly moved from their external-merge position, as in (43) and (44). Unmoved contrastive foci are not ungrammatical, but they are slightly unnatural. In this respect, it is possible that Trevisan speakers are influenced by standard Italian in their judgements, given that the language allows contrastive foci both in situ and fronted to the HLP (see Bianchi 2013 and related).

The first hypothesis suggested by the movement in (43b) and (44b) is that Trevisan corrective foci target a low focal projection in the LLP, Belletti’s (2004) Foc, as in (45):

Significantly, total fronting of contrastive foci is marginal in Trevisan and, in non-matrix environments, it requires the realisation of a co-indexed clause-internal clitic (when available), i.e., it is further topicalised into the HLP. Observe the contrast between a fronted and a clause-internal focus in (46):

What constructions like (46) suggest is that the functional projection relevant to focus of any type is always FocP in Trevisan, not Rizzi’s (1997) left-peripheral FocusP: foci naturally move to the low peripheral focus position and, to be fronted into the HLP, require further topicalisation.

Horvath (1986) argued that whenever a language displays a specialised projection for contrastively-focused constituents, the same projection is also available for wh-elements. Horvath explained this property on the basis of the interpretational similarities displayed by focus and wh-elements: in contrast to informational focus, contrastive focus operates over a closed set and, in a similar fashion, the value of wh-elements is drawn from an inferable (hence closed) set of items, inherently delimited by the truth value of the question itself. An extension of the movement paradigm sketched in (45) to clause-internal wh-elements is thus straightforward, and appears justified semantically. That Trevisan contrastive foci make use of the LLP constitutes evidence that clause-internal wh-elements do undergo focus movement in this language, as in the non-Romance languages overviewed in (§2.3).

This low movement of wh-elements, which I shall henceforth refer to as Wh-to-Foc, has non negligible consequences for the theory of wh-in situ, as I discuss in what follows.

Cable (2010) argued that, for numerous phenomena surrounding wh-operators, the locus of explanation does not lie in wh-operators themselves, but rather a distinct element bearing a special semantic and/or syntactic relationship to the wh-operator, namely the Q-particle. Despite significant disagreement over many issues related to interrogative wh-movement, the literature indeed exhibits a common consensus: that the interrogative fronting of wh-elements directly results from a property borne by the wh-element itself, and that wh-questions where there is fronting of a phrase reveal the existence of pied-piping. To Cable’s understanding, the wh-questions of Tlingit strongly challenge this classic stance, and a model is required where wh-fronting is not triggered by any properties of the wh-element but rather targets the features of the Q-particle. This treatment has important consequences for the general theory of wh-in situ.

In Tlingit wh-questions, the wh-element precedes the main predicate, and is typically clause-initial. The wh-element is construed with the Q-particle sá, which directly follows either the wh-element or a larger phrase that contains it. The remaining material in the sentence follows the wh-element, as illustrated in (47):

Cable argues that wh-fronting is an instance of fronting of the Q-particle that results in somewhat parasitic pied-piping of the wh-element. In his account, fronted wh-elements have the composite structure in (48):

Accordingly, in Tlingit the Q-particle sá takes its sister as a complement, with the result that a QP node immediately dominates both the Q-particle and its sister, the wh-element. Attraction of the Q-feature to the HLP thus entails that the entire QP is moved overtly, as in (49) (throughout, dashed arrows are used for covert movement and Agree relations):

The analysis in (49) is empirically supported by the fact that the felicity of a Tlingit wh-question depends only upon the locality of the Q-particle to the HLP, while the locality of the wh-element appears to be irrelevant. Observe for instance the islands in (50):

In a Tlingit wh-question, the wh-element can surface inside an island iff the Q-particle surfaces outside of the island, as in (50a). If the Q-particle is located inside the island, as in (50b), the sentence is ungrammatical. The pattern in (50) argues that the rules for forming wh-questions are only sensitive to the surface position of the Q-particle, meaning that only the features of the Q-particle can be referenced by those rules. Therefore, Cable suggests that what is usually understood as ‘pied-piping’ of wh-elements is rather an instance of QP-fronting that results in parasitic movement of the wh-element to the HLP. This, for Cable, is true of all languages, including those which lack phonetically-realised Q-particles.

Based on the understanding of overt QP-fronting sketched here in (49), Cable formulates a new typology of wh-in situ whereby wh-in situ languages comprise at least two distinct syntactic types:

In Q-adjunction languages, the Q-particle and the wh-elements are sisters. The node which immediately dominates the Q-particle and the wh-element is not a QP, but rather of the same type as the wh-element, as outlined in (51):11

For Q-adjunction languages, Cable adopts Hagstrom’s (1998) treatment of Japanese wh-questions, along the lines of the diagram in (52):

In the presence of Q-adjunction, attraction by the Q-feature into the HLP entails only that the Q-particle moves while the wh-element is stranded clause-internally. In languages of this type, the Q-particle is therefore fronted to the HLP alone, as in the Japanese and Korean examples in (53) and (54):

In contrast, in wh-in situ languages with Q-projection, the structure of the wh-element is the same as that previously posited for Tlingit. The only difference with respect to Tlingit lies in the timing of movement, which takes place in covert syntax here, as in (55):

An example from SOV Sinhala is provided in (56):

In what follows, I argue that an extension of the theory elaborated by Cable to Romance, coupled with the predictions of Wh-to-Foc, allows the formulation of a theory of wh-in situ which is anchored on three theoretical pillars that are central to any investigation done within any framework of generative grammar: Universality (the assumption that structures are not language-specific but rather fixed across languages); Uniformity (Chomsky’s 2001 invitation to explain cross-linguistic variations as restricted to easily detectable properties of utterances); and Economy (the attempt to explain languages using structures which are as uncomplicated and learner-friendly as possible). The result will be a universal yet uncomplicated answer to the composite phenomenon of (pure and optional) wh-in situ in terms of different combinations of micro-parametric settings.

Cable’s proposal for a cross-linguistic implementation of Q-particles in answer-seeking wh-interrogatives is perfectly compatible with the work of the cartographic enterprise, into which the present article falls. If the existence of one functional head in one single language is enough to posit the existence of such head in all languages, then the existence of phonetically-realised Q-particles in many languages must be acknowledged and posited cross-linguistically.

Remember that Trevisan allows both wh-fronting into the HLP and low focus movement into the LLP, as seen in (3) and (4). On Cable’s assumptions discussed in §3.1, this type of alternation can only be explained if Trevisan is a mixed Q-projection/Q-adjunction language with silent Q-particles: accordingly, wh-fronting is parasitic on QP-fronting, while Wh-to-Foc results from the presence of Q-adjoining structures (see §3.4 for details). That the Trevisan alternation between high and low wh-fronting is not derived with Q-projection alone coupled with an alternation in the timing of movement, i.e., overt QP-movement vs covert QP-movement, is witnessed by the fact that SCLI signals the presence of overt V-to-C movement in the interrogatives of Trevisan.

A comparison of the movement properties of wh-elements and foci in Trevisan and Italian sheds light on the loci where the features relative to interrogative movement are encoded in the functional spine, with intriguing repercussions for the theory of wh-in situ.

In languages like standard Italian, contrastively focused constituents naturally occupy a low position, where the sentence stress falls. However, these can also move to a left-peripheral position, where they bear a particular pitch accent, as illustrated in (61):

It is commonly assumed that focus in the low position, as in (61b), can either carry new information or be used in contrastive contexts, while the high position in (61c) can only be used contrastively. Regardless of the reasons behind this apparent optionality, the functional projection that encodes [foc] in standard Italian is a left-peripheral one, since clause-internally moved foci and wh-elements are not licensed, as in (62) and (63):

Significantly, while the total fronting of contrastive foci can be delayed to LF in standard Italian, as illustrated in (62b), wh-fronting must be done overtly, along the lines of (64):

Since Rizzi (1997), much work has taken focus-fronting and wh-fronting to compete for the same left-peripheral specifier, SpecFocusP. This is argued to justify the complementary distribution of the two constructions, at least on syntactic grounds. Given the semantic and syntactic similarities between the two constructions, it is legitimate to wonder why standard Italian can delay focus-fronting until LF, while covert wh-fronting is systematically disallowed. It is possible to explain this alternation as another intermediate diachronic stage, i.e., that Italian is slowly acquiring the possibility to encode [focus] in the LLP (see also the discussion of Italian wh-in situ in Badan & Crocco 2021).

Given Cable’s claim that all instances of fronting of wh-elements are triggered by [q], coupled with the observation that interrogative words in wh-interrogatives are inherently focused, and hence carry both [q]- and [foc]-features (encoded by the the Q-particle and the wh-element, respectively), the observed differences between Trevisan and standard Italian can be explained in terms of the loci where the two features are realized in the functional spine: while Trevisan ‘scatters’ the two features between the HLP and the LLP (in the sense of Giorgi & Pianesi 1997), standard Italian realises them as a bundle in the HLP. This is illustrated in Table 1:

In Table 1 and throughout, I use the EPP as a formal requirement for overt movement.

Let me now outline the computations responsible for Wh-to-Foc and QP-fronting in Trevisan, and then survey some broader empirical predictions of my theory.

Considering my discussion of the Trevisan facts, I wish to suggest that the projection parameter ought to be modified as follows:

The major consequence of the parameter above is that only one kind of total fronting exists, namely that of Q-projection languages that move QP overtly. According to Cable, the existence of QPs should also be posited in languages with silent Q-particles. Here, I have argued that Trevisan SCLI supports Cable’s claim empirically. In contrast to Cable (2010) though, I have claimed that there are languages with both Q-projection and Q-adjunction, a property that explains the apparent optional in situ/ex situ alternation of Trevisan and more generally NIDs, and is supported by the interrogative morphosyntax of Ancash Quechua. On these assumptions, the fact that in some ‘optional in situ languages’ not all wh-elements can surface either clause-internally or sentence-initially (see, for instance, Munaro’s 1999 discussion of Pagotto) can be assumed to be related to: (i) either special properties of the wh-elements under consideration (for example, why-words externally-merged directly in the HLP) or (ii) the fact that, in intermediate stages in linguistic evolution, it is unsurprising for Q-projection and Q-adjunction not to apply to all types of wh-elements homogenously.

Another prediction of Cable’s approach is that there can be basically three types of wh-in situ languages: (a) Q-projection languages that move QP covertly, (b) Q-adjunction languages that move Q covertly, and (c) Q-adjunction languages that move Q overtly. I have argued that the apparent optionality of Trevisan (and Romance in general) can be better explained if it is assumed to derive from the exceptional existence of both Q-projection (responsible for total fronting) and Q-adjunction (responsible for wh-in situ). Because of the presence of SCLI in Trevisan, both in constructions with QP-fronting and with wh-in situ, I have claimed that this language is of the (c) type, and that it has an EPP-feature both in the LLP and in the HLP.

I also identified a special type of movement of clause-internal wh-elements that has not been discussed by Cable, which I call Wh-to-Foc. This movement is linked to the need to check a [foc]-feature in the LLP, and is then triggered by an EPP-feature. Considering the theory of Wh-to-Foc developed here, an additional parameter ought to be added to Cable’s:

In languages of the ‘scattering’ type, low movement of wh-elements is observed in case the language has an EPP feature in the LLP.

In what follows, I suggest possible evolutionary paths that the grammar of Q, coupled with my discussion and some relevant diachronic data, predicts for the cross-linguistic morphosyntax of single wh-questions.

Works on pure wh-in situ illustrate that there exist languages which underwent interesting typological changes. For instance, Watanabe (2003) claimed that Japanese, which is today a Q-adjoining language, went from overt wh-fronting into the HLP during the Nara period (8^th century) to unmoved wh-in situ. Examples of wh-fronting in Old Japanese are provided in (76):

Considering Cable’s (2010) assumption that wh-fronting is systematically QP-fronting, (76) could be taken as proof that the linguistic evolution goes from QP-fronting to unmoved Q-adjunction, passing through a phase characterised by optionality, such as the one observed today in Romance. Significantly, Watanabe claimed that, in Japanese, wh-fronting co-existed with wh-in situ in the Heian period (9^th to 12^th century).

A different evolutionary path is suggested by Aldridge’s (2009) analysis of the Old Japanese examples in (76). According to her, Watanabe’s claim is partly based on his assumption that genitive subjects are in the canonical subject position, and hence he analyses a preceding wh-element as having been raised out of TP. In contrast, for Aldridge genitive subjects do not exhibit the behaviour of nominative subjects in SpecTP, and their distribution is better understood if we assume that they do not move from SpecvP. If the genitive subject occupies a very low position, then a Wh-to-Foc movement analysis becomes available for clause-internal wh-elements such as those in (76). It is therefore possible that Japanese was never a QP-language, and Old Japanese should rather be analysed as a Q-adjunction language, like the contemporary variety. Accordingly, the presence of an EPP-feature in the LLP, no longer present today, triggered overt movement in Old Japanese.

Similarly, Aldridge’s (2010) work on Archaic Chinese (Warring States period, 5^th to 3^rd century BCE) suggests that Chinese went from having clause-internally moved wh-elements, as shown in (77), to present day unmoved wh-in situ:

Aldridge’s works therefore suggest that both Japanese and Chinese moved from what looks like Trevisan-style Wh-to-Foc to modern-day unmoved in situ.

Based on the evolutionary patterns that I have sketched here, I wish to suggest that the evolution of wh-interrogatives goes from overt QP-fronting to either covert QP-fronting or unmoved Q-adjunction. I discussed these cases here to open my analysis of Trevisan to the investigation of other optional and pure in situ languages. Indeed, these patterns do constitute an interesting starting point for future research on single wh-interrogatives.