3.1 Clause-internal wh-phrases

The different linear orders attested in NIDs have been discussed in the literature as resulting from three main strategies: (i) Remnant movement; (ii) In-situness; and (iii) Short movement. We briefly discuss these accounts below (see also Bonan 2019; 2021a; Fiorini 2021; Neagu 2022; Munaro & Poletto 2023).

3.2 Wh-doubling

In all accounts discussing doubling in NIDs (see Munaro & Poletto 2023: for an overview), the linearly-first wh-phrase is considered to be part of the left periphery of the clause. The lower wh-phrase is discussed by the proponents of the remnant movement analyses as occupying the lower WhP or an OpP position identified for clause-internal wh-phrases, also in the CP field.

Poletto & Pollock (2004b) analyze wh-doubling as the A-bar counterpart of pronominal clitic doubling, common across Romance languages. They propose that the wh-phrases are (e-)merged as a complex Cl(itic)P, with the preverbal element heading the projection and the postverbal one occupying the specifier position (23):

They postulate this configuration because the preverbal element is always morpho-phonologically smaller than the lower one, which is taken to be evidence of a clitic-like status.

Besides the absence of further evidence for this claim, as mentioned above, Manzini & Savoia (2011) notice that in Lombard varieties, the preverbal element may have a longer form (24a), and it can be indeed the sole wh-phrase present in fronting structures (24b)

According to Poletto & Pollock (2004b), the clitic wh-phrase raises to IP into its cliticization site before moving to the left periphery. However, as the authors notice, undergoing movement into its final landing site would mean crossing heads. The subject-clitic inversion (SCLI), which is present in all direct questions either in the form of an enclitic attached to the end of the finite verb or in the form of a free subject clitic, comes in handy since it would drag along the clitic wh-phrase to the left periphery.

In addition to the lack of evidence for such movement, SCLI is not present in embedded environments, meaning that a clitic wh-phrase should not be able to reach its landing site.⁹

Consider (25), which shows that the wh-phrase is the only element surfacing in a left-peripheral position:

Furthermore, the wh-clitic movement would also violate the Head Movement Constraint (HMC: Roberts 2018), as it must skip the subject clitic head to reach the final landing site in the left periphery.

As stated in the previous section, Bonan (2021b) provides an analysis relying on the concept of low periphery of the clause for the lower wh-item. The higher element, on the other hand, is in CP. Bonan, building on Cable (2010), proposes that optionality between the three strategies for whPs in NIDs is the byproduct of two distinct lexical strategies, one involving QP-selection, and one involving QP-adjunction, as depicted in (26):

According to the author, optionality arises when the [Q] and [foc] features on wh-phrases are checked according to different strategies: some languages check both features in C, yielding wh-fronting, while others check [foc] clause-internally, yielding a wh-phrase in the low periphery. In other words, Bonan’s proposal for the fronted vs clause-internal configurations relies on the concepts of a Q-operator in CP (Cable, 2010), and an optional [foc] feature that needs to be checked in the left periphery of vP. The latter is the approach that most resembles our own; however, while we maintain, also in line with Manzini & Savoia (2011), that the two elements possess distinct features, we also diverge from these analyses by proposing a different syntactic operation as well as a different distribution of formal features on the wh-items.

In Section 4, we agree with the first part of the analyzes of Manzini & Savoia (2011) and Bonan (2021b) showing that the dislocated element is indeed a pure operator. However, the second part seems to be better explained on morpho-phonological grounds. At LF, we analyze the clause-internal wh-phrase as the one bearing the wh-feature, without recurring to focal properties, which would contradict scope-related considerations made for the operator portion of the wh-chain.

4.1 Doubling

Doubling interrogative constructions are attested in Monnese, but not in Darfense. In the former variety, these interrogatives are only possible with a subset of wh-pronouns: what and where. There is a rigid distribution of the two wh-elements forming the doubling constructions, such as in (27) and (28):

The two structures in (27a) and (28a) also exemplify the word order in doubling constructions. The first element in the interrogative is the leftmost wh-element, followed by the auxiliary verb and its subject clitic. The subject precedes the lexical verb and the right-adjacent wh-item.

This word-order provides a first indication of the positions occupied by the wh-elements: the edge of the CP and vP phases. Interrogatives in Camuno are characterized by subject-clitic inversion (SClI), as observable in (27a) and (28a), where the auxiliary verbs are followed by their subject clitic in linear order. In SClI structures, the inflected verb has been argued to raise to C (see Rizzi & Roberts, 1989 for French; and Poletto, 1993 for NIDs), meaning that the fronted wh-item occupies a position within this functional field, i.e., either the/a specifier of CP, or FocP in the left periphery (Rizzi 1997). We will assume, by adopting a more minimalistic approach, that the fronted wh-element lands in spec,CP. As far as clause-internal wh-elements are concerned, we show that they are hosted in a position at the edge of vP, rather than in situ. Consider (29):

Example (29) shows that when a manner adverb adjoining to VP like d’ nse ‘ngurt ‘so greedily’ is present, the linear order maintains the right-adjacency of the wh-element to the lexical verb, and the adverbial follows the wh-phrase. Furthermore, in ditransitive structures, the object must follow the dative wh-phrase (30), or viceversa, in the case where the wh-item is the direct object:

The one in (30) does not correspond to the pattern visible in declarative structures, where the direct and the indirect objects can be freely reordered (31):

Past participle raising has been attested in multiple Romance languages (see Cinque 1999; Alboiu 2003), a phenomenon which also applies to Camuno, as illustrated in the following example (32). The past participle in (32) is shown to raise from VP over a frequency adverb located in the inflectional domain, further indicating that the clause-internal wh-element surfaces at the vP phase edge and not lower/in situ:

Based on the distribution of the two wh-phrases, we analyse them as occupying parallel positions at the CP and the vP phase edges. We therefore propose an analysis in terms of partial copying, which has been applied by Barbiers (2008); Barbiers et al. (2010) for non-identical doubling in Dutch. However, we only adopt the core idea of partial copying as an operation targeting a specific node of a layered projection, also considering the differences between Dutch and Camuno doubling interrogatives.¹⁰

The analysis we propose for the doubling constructions is summarized in (33) and represented in (34):

In terms of formal features, we propose that the complex XP that we just introduced is formed by two elements bearing two distinct formal features: on one hand, we have QP that carries the uOP feature, and on the other hand we have whP which carries the iwh-feature. Put differently, the literature has shown that wh-phrases are characterized by two main formal features: an operator feature and a wh-feature (see Canac Marquis & Tremblay 1998; Den Dikken 2003). In the case of Camuno, however, these features are lexicalized in two distinct positions at the phase edges: the operator feature is spelled-out in CP, and the wh-feature is spelled-out at the vP edge. Moreover, in order for partial copying to apply successfully, we propose that the uOP* feature on QP is strong, obligatorily raising to CP, while the uwh feature in the Camuno interrogative C head is weak, which allows for long-distance Agree between C and whP at the edge of the lower phase. This also explains why the clause-internal wh-element occurs at the periphery of vP, instead of in-situ: the first, short movement of the whP cluster to this position is caused by the Edge Feature in vP. This feature allows for the complex whP to escape the lower phase before this is sent to the interfaces, thus allowing for feature-checking operations between the wh-elements and CP to take place. Finally, we explain the optionality between the three strategies available in the presence of doubling by relying on the LF and PF interfaces. When only one wh-item is lexicalized, it means that it is interpreted at LF, but does not have a realization at PF.

In other words, the derivation in (33) applies to all types of interrogatives under investigation: wh-fronting, clause-internal whP, and wh-doubling. The difference between them is due to only one wh-item being pronounced at PF in the former two (the operator in the case of fronting and the element carrying the wh-feature in the case of clause-internal whP) and both wh-items being pronounced in the latter.

Evidence supporting this analysis comes from other A-bar dependencies and intervention effects, as well as syntactic theories dealing with successive cyclicity. Firstly, the fact that doubling is non-identical and that the fronted element cannot appear clause-internally or vice versa, indicates that each wh-item plays a specific role within the clause, lexicalizing different formal features. In the same vein, the fact that the fronted wh-element is used as an operator in other types of constructions indicates that this is indeed the operator portion of the chain. See the following restrictive (35a) and free (35b) relative clauses:

Secondly, intervention effects and sensitivity to islands indicate that the raised element and the stranded one are part of the same A-bar chain. Indeed, if this analysis is on the right track, we would expect sensitivity to islands in a situation where there is an intervening operator between the operator feature and the wh-feature in an A-bar chain. The role of intervening operators in A-bar chains has been in fact thoroughly investigated in early studies on island effects (see Cinque 1990; Rizzi 1991: i.a.). Another thorough discussion on intervention effects in A-bar dependencies has been provided by Beck (1996) and Miyagawa (2004). These constraints, we argue, are the source of the ungrammaticality of (36) below:

In (36), the negative operator miga blocks AGREE between CP and the lower wh-item bearing the wh-feature, and disrupts the A-bar chain formed by the two distinct wh-elements under analysis.

The only way to form a negative interrogative in Camuno is via clefting, as illustrated in (37):

Finally, our proposal is consistent with the hypothesis for phasal successive-cyclicity for A-bar movement (Chomsky 2001; 2009; Gallego 2012; van Urk 2020), in particular providing evidence for the edge of the vP phase as an intermediate step. Incidentally, since the pattern remains consistent with all types of structures, i.e., including passives and unaccusative ones, our derivation also shows that all vPs may indeed be phases (in line with Legate 2003; 2005). As mentioned above, the exact position of the clause-internal wh-element depends on the framework one adopts. In a more strict version of minimalism, this position could be one of the specifiers of vP, used as an escape hatch for elements that must not be transferred to the interfaces in the lower phase, as mentioned earlier in this section. In a cartographic framework, the lower wh-item could be analyzed as occupying a functional projection in the low periphery, e.g. lower FocusP (Belletti 2004). We prefer to adopt the former approach in this case.

The following section will look better at the clause-internal wh-phrase in the absence of an overt operator in the CP field, as observable in most Camuno varieties.

4.2 Clause-internal distribution

Interrogative constructions with a clause-internal wh-phrase are the most common in Camuno varieties. We analyze this element as a lexicalized wh-feature in an A-bar dependency whose operator in spec,CP is silent. Its position within the clause is the same as the one seen in the doubling interrogatives described above. See (38) and (39):

Example (38) shows that wh-items must be followed by subjects, and not vice versa.¹² At the same time, the distribution of the lower wh-phrase in ditransitives mirrors the one observed in doubling interrogatives (39). We take this to indicate that the clause-internal wh-element indeed covers the same role in the A-bar chain as the one described for doubling wh-phrases, located at the edge of the vP phase. We adopt the same analysis for this element because clause-internal interrogatives exhibit the same intervention effects observed in doubling constructions. Hence, we assume a silent operator in CP (40):

The clause-internal wh-phrase in (40a) is sensitive to islands, which we take to indicate that the negative operator is an intervening barrier disrupting the A-bar chain between the wh-feature bearing item and a silent operator located in the CP field. As shown for doubling constructions, the only way to circumvent this is by employing a cleft structure (40b).

4.3 A unified account

This section provides a unified account for the variation observed in Camuno interrogatives, i.e., the presence of the two distinct strategies introduced in the previous sections. Providing a unified account for constructions exhibiting this amount of variation and optionality is no easy task. For this reason, even though this article is dedicated to a theoretical syntactic analysis of wh-interrogatives in the aforementioned dialect, we believe that the heterogeneous behavior of the different whPs analyzed throughout section 4 (and the reason why some of them are only touched upon in our analysis) is related to language evolution and diglossia.¹³ In particular, it has been shown that, similarly to free variation in phonology, competing grammars within the same individual/community of speakers give rise to optionality as a result of speakers employing two or more grammars (Lightfoot 1999; Kopf & Weber 2023), in this case, Camuno and standard Italian. Moreover, the presence of two competing grammars could also explain why only certain wh-categories are able to be doubled, as well as why fronting, with the exception of why, tends to be limited to formal topics of conversation. Put differently, we believe that we are currently observing a transitional stage in Camuno evolution, where standard Italian has been slowly replacing local languages and varieties not only in formal contexts, but also in informal ones.

Moving back to the syntactic analysis, we have already mentioned that the lower element in wh-doubling interrogatives is the same, in the same position, as the one of clause-internal/postverbal one. A way to account for the data is by looking at the relationship between syntax and its interfaces. Specifically, in wh-doubling constructions, both elements of the wh-chain are phonologically realized, meaning that the information transferred from narrow syntax to LF, is also transferred at PF. On the other hand, in those varieties where only the lower element is spelled out (or when this is the absolute preferred strategy, even when fronting may be an available option), e.g., Darfense, the operator portion is interpreted at LF, but is not overtly realized at PF. This approach could also explain optionality in Monnese: since we do not have any evidence that optionality is triggered by extra grammatical factors such as different pragmatic readings, we assume that it is an interface-related phenomenon. Thus, Monnese may optionally spell out either the fronted or the clause-internal wh-item, as well as both these elements. Whenever only one of the wh-items is phonologically realized, however, the other is still interpreted at LF.

This analysis may be problematic if one notices that the operator of other A-bar dependencies in Darfense has the same form as the element we analyze as the one carrying the wh-feature in the interrogative constructions, as exemplified in (41).

However, non-identical doubling is not the only pattern found in Eastern Lombard dialects like Camuno. Indeed, identical doubling is also attested in these dialects, as illustrated by Manzini & Savoia (2011). See (42):

We suggest that, in those varieties and constructions that we analyze as involving a silent operator, the LF structure may resemble the one illustrated in (42). This would explain why the lower wh-element has the same form as the operator in other A-bar dependencies.

The reason for the presence of structures with an overtly realized clause-internal wh-item and a silent operator is, once again, related to the relationship between syntax and its interfaces. First of all, we take the lower wh-element as occupying a position at the lower phase edge. Phases are cyclic syntactic domains, constituting points of transfer to the interfaces with PF and LF (see Chomsky 2001). Secondly, we will better show in the following section that the main sentence stress in Camuno generally aligns with the right edge of the predicate phonological phrase. This is the portion of the clause in which clause-internal wh-elements surface. Based on these observations, we conclude that the wh-feature is realized overtly instead of the operator because the edge of the vP phase in Camuno is more articulated. This analysis of a highly articulated lower phase is supported by data involving wh-doubling in long-distance dependencies, like in (43):

As it can be observed in (43a), the intermediate step where the lower wh-item is stranded in a long-distance dependency is at the edge of vP in the embedded clause. Languages such as Dutch and German, that display evidence for successive cyclicity in an intermediate step in long-distance A-bar movement, mark the CP phase of the embedded clause as an intermediate position. The same does not apply to Camuno, where the lower wh-item cannot surface in Spec,CP, as shown in the ungrammatical sentence in (43b). Instead, it must surface at the periphery of vP. We interpret these data as syntactic evidence for a preference of the vP phase edge over CP in Camuno. The next section will better account for this observation by looking at the PF interface.

4.3.1 PF interface

In this section, we formulate a tentative hypothesis rooted in the properties related to stress assignment in Camuno to explain the preference for the postverbal distribution of wh-phrases in varieties where doubling is not available. Furthermore, we suggest that the same factors affect the morphological form of the WhPs in doubling structures, resulting in the asymmetry discussed above. In particular, given the unified derivation proposed in the previous section, we attempt to explain why the pronunciation of the lower element (the whP) is preferred over the one of the fronted operator.¹⁴ Our proposal is not new in its main idea. The hypothesis of a direct correlation between marking and interpretation of informationally marked items (in particular foci) and the way structures are mapped at PF has been put forward in the literature on Italian and Italian varieties over the years (see Frascarelli 2000; 2012; Szendrői 2001; 2017; Bocci 2008; Féry 2013; Bocci & Avesani 2015; Samek-Lodovici 2015; Bocci et al. 2024: among others). This literature provides both theoretical and experimental evidence supporting the main claim that information structure is not relegated to a discourse-pragmatics component, but it directly affects (or it is affected by), operations at the so-called syntax-prosody interface.

4.3.1.1 Stress distribution and information structure in Camuno

In unmarked, out of the blue structures, the main sentence stress in Camuno aligns with the right edge of the utterance in (44), as shown in Figure 1:

1. (44)

1. i’
2. cl.3pl
1. a
2. have.3pl.prs
1. iht
2. see.prt
1. al
2. the
1. hpetacol
2. show
1. co
2. with
1. le
2. the
1. ho
2. their
1. horele.
2. sisters
1. ‘They saw the show with their sisters up.’

Figure 1: Pitch contour of i’ a iht al hpetacol co le ho horele “They saw a show with their sisters.”

In marked structures, stress aligns with the most informationally prominent element in the sentence which, like the clause-internal wh-phrase, is generally hosted in the postverbal position. These elements include all types of foci, emphatic negative particles, discourse particles, and discourse markers.

The configuration is not typologically unknown, and it has been extensively discussed, for instance, in the literature on Bantu languages (Aboh 2007; Zerbian 2007; Hyman & Polinsky 2009; Buell 2015), Basque (Elordieta 2001; Arregi 2002; Fiorini 2023), and Italian, Gallo-Italic, and Italo-Romance varieties (Samek-Lodovici 2015; Szendrői 2017; Bonan 2019; Neagu 2022; Fiorini 2024b).

Cross-linguistically, informationally salient elements are associated with prosodic prominence. In Camuno, these are hosted in the postverbal position and include foci and wh-phrases.¹⁵ The definition of prominence that we follow is that in Himmelmann & Primus (2015), who propose that prominence is relational, dynamic, and that it attracts linguistic operations. From a semantic point of view, a wh-interrogative is prominent because it activates alternatives (Rooth 1985). This characterization is generally discussed in the literature of focus, but questions do not necessarily exhibit all its properties and we thus follow, among others, Rizzi (2001); Poletto & Pollock (2004a); Rooth (2016) in considering questions a separate semantic object. In syntax, this corresponds to the presence of the characteristic [Q], rather than [foc]. From a phonological point of view, the relative prosodic salience of the target element and the environment in which it appears affects the way hearers interpret the informational load of the relevant proposition (Baumann & Schumacher 2020; Grice & Kügler 2021; Ladd & Arvaniti 2023).

This is visible in Figure 2, which represents the pitch contour for (45), a structure in which the corrective focus a Bré ‘in Breno’ surfaces in postverbal position:

1. (45)

1. (no,)
2.   no
1. i’
2. cl.3pl
1. a
2. have.3sg.prs
1. iht
2. see.prt
1. a
2. at
1. Bré
2. Breno
1. al
2. the
1. hpetacol
2. show
1. co
2. with
1. le
2. the
1. ho
2. their
1. horele.
2. sisters
1. ‘(No,) It is in Breno that they saw a show with their sisters’

Figure 2: Pitch contour of i’ a iht a Bré al hpetacol co le ho horele ‘It’s in Breno that they saw a show with their sisters (not in Bienno).’

Focus does not necessarily align with the highest pitch in the utterance, albeit remaining prosodically prominent. This is the case of structures containing contrastive topics like the one in (46), represented in Figure 3. Here, the subject Piero aligns with the main sentence stress, which is followed by an intonational break, arguably creating a specific prosodic domain (see , in particular, Féry 2007). In this type of structures, the focus tre ‘three’ carries a marked but less prominent stress that results in the desired interpretation:¹⁶

1. (46)
1. a.

1. i
2. the
1. do
2. two
1. fioi
2. kids
1. de
2. of
1. Piero
2. Piero
1. i
2. cl.3pl
1. laura
2. work.3pl.prs
1. a
2. at
1. Loer.
2. Lovere
1. ‘Piero’s two sons work in Lovere.’

1.  
1. b.

1. pota
2. pota
1. ma
2. but
1. Piero
2. Piero
1. ge
2. cl.3sg
1. n’
2. cl.loc
1. a
2. have.3sg.prs
1. tre
2. three
1. de
2. of
1. fioi!
2. sons
1. ‘But Piero has THREE sons!’

Figure 3: Pitch contour of pota ma Piero ge n’ a tre de fioi! “But Piero has THREE kids!”.

As for interrogative structures, in polar questions the constituent inquired about (i.e., the most informationally prominent) appears postverbally, and aligns with the main sentence stress (47) as shown in Figure 4:

1. (47)

1. abitef
2. live.2pl.prs
1. a
2. at
1. Darf
2. Darfo
1. oter?
2. you
1. ‘Do you all live in Darfo?’

Figure 4: Pitch contour of abitef a Darf oter ‘do you all live in Darfo?’.

The pattern holds for wh-phrases, which, as discussed above, are generally argued to belong to the group of alternative-evoking elements that require marking (Rooth 2016). Consider (48) and its intonational contour in Figure 5:

1. (48)

1. a-i
2. have.3sg.prs=cl.3pl
1. parlat
2. talk.prt
1. de
2. of
1. ki
2. who
1. anco
2. today
1. al
2. a.the
1. telegiornale?
2. news
1. ‘Whom did they talk about on the news today?’

Figure 5: Pitch contour of ai parlat de ki anco al telegiornale ‘Whom did they talk about on the news today?’.

As expected, in Figure 5, the main sentence stress aligns with the wh-phrase de ki ’of whom’, which surfaces postverbally.

4.3.1.2 Derivation of wh-interrogative prosodic structure

A clear pattern emerges: in out-of-the-blue, unmarked structures, the main sentence stress in Camuno aligns with the right edge of the utterance. In informationally-marked structures, stress aligns with the relevant element which is hosted in the postverbal position. In other words, there seems to be a fundamental friction between two phonological constraints. Such a contrast has been discussed in these terms by, among others, Féry (2013), who identifies in the focus-stress correspondence a condition to be met by aligning the latter with the right edge of the phonological phrase containing the focus. That is, stress falls as close as possible to the right edge of the utterance, while marking the appropriate constituent. A similar account is proposed in Szendrői (2017), but in her account it is the phrase containing the verb to count. Considering the distribution of both unmarked and marked prominence, we propose that the same operation occurs in Camuno for postverbal wh-phrases, which share the distribution of foci.¹⁷

In (49), it is represented the simplified prosodic structure of an utterance in Camuno. Stress, marked in red, falls on the phonological word ω at the right edge of the phonological phrase φ, which is, in turn, at the right edge of the intonational phrase ι. In other words, stress aligns with the right edge of ι.

1. (49)

1. 

Consider now (50b) and (50c), which represent the prosodic structure of a wh-interrogative like (50a). In this case, stress falls on the rightmost word (the wh-phrase) of the leftmost (in this representation) phonological phrase:

1. (50)
1. a.

1. a=l
2. has=cl.3sg
1. majat
2. eat.prt
1. kwando
2. when
1. al
2. the
1. Piero
2. Piero
1. la
2. the
1. polenta?
2. polenta?
1. ‘When did Piero eat polenta?’

1.  
1. b.

1. [[a=l majat KWANDO]_φ[al Piero la polenta] _φ]_ι

1.  
1. c.

1. 

Following standard assumptions, we propose a mapping algorithm to build the phonological structure of a wh-phrase based on a cyclic application of the Nuclear Stress Rule (Cinque 1993; Arregi 2002; Zubizarreta & Vergnaud 2017) rooted in the Multiple Spell-Out model (Chomsky 1995; 2001; Uriagereka 2011). The former assigns the main sentence stress to the most embedded phrase in the syntactic tree. In particular, we follow Arregi’s 2002 version, for which stress is assigned to the most embedded branching node.

Crucially, here we assume that the prosodic structure of any sentence is built incrementally by phases, in line with a basic minimalist assumption (see Chomsky 1995; 2001). That is, when a phase is completed and transferred to the interfaces, it creates a phonological phrase (Kahnemuyipour 2009) that represents a viable stress domain (Reinhart 1995; 2006; Féry 2013; Szendrői 2017).¹⁸ According to our analysis, the lower wh-phrase (i-)merges at the edge of the vP phase. That is, it is spelled out as the element most embedded in the phasal domain C, representing the right edge of a phonological phrase. It can thus align with main sentence stress, which is selected at PF. Based on experimental evidence, Bocci et al. (2021; 2024) show that stress assignment reflects the steps of the syntactic derivation (short vs. long movement) of wh-structures in Italian. This is in line with the cyclic structure-building operation describe above, and with the availability of multiple potential loci for stress assignment.

This syntax-prosody mapping operation can straightforwardly explain the different forms and their distribution on phonological grounds, departing from the assumption of a bipartite categorization to account for the two morphological forms for whPs in several NID, including Camuno (contra Poletto & Pollock 2005; Manzini & Savoia 2005; Bocci et al. 2022). The evidence supporting such a hypothesis can be accounted for as a byproduct of the restriction on PF. Consider the examples in (27) and (28), adapted and reported here as (51):

1. (51)
1. a.

1. k’
2. what
1. e-t
2. have.2sg.prs=cl.2sg
1. sumena
2. sown
1. kwè?
2. what
1. ‘What did you sow?’

1.  
1. b.

1. ‘ngo
2. where
1. se-t
2. be.3sg.prs=cl.2sg
1. nasciut
2. born
1. ‘ngont?
2. where
1. ‘Where were you born?’

The two wh-phrases are strictly ordered, and the lower one is invariably the phonologically heavier one.¹⁹ We argue that their realization is directly affected by the stress pattern which favors stressing on the right element in a phase (i.e., an intonational unit).

In structures where only one wh-phrase is pronounced, the economy principle favors the application of a mechanism similar to “pronounce low copy (PLC)” (Bošković 2001), which can apply to A’-chains (Seguin 2024). Prominence is assigned to a heavy element close (or at least “as close as possible” (Féry 2013)), to the default position for stress (i.e., the rightmost edge of the phonological unit containing the whP).

Assuming an important role of phonological restrictions on the morphological form of wh-phrases can also provide a tentative explanation about the reason for which fronted wh-phrases in the rare case of “simple fronting” distibution are consistently phonologically heavier. In order to resist PLC and be pronounced as the only interrogative element when lexicalized, the operator needs to attract stress, and it is thus lexicalized in a way that outweighs the less costly rightward stress assignment (52):²⁰

1. (52)
1. a.

1.   e-t
2.   has.cl.3sg
1. fat
2. do.prt
1. kè?
2. what

1.  
1. b.

1. *kè
2.   what
1. e-t
2. has.cl.3sg
1. fat
2. do.prt

1.  
1. c.

1.   koha
2.   what
1. e-t
2. has.cl.3sg
1. fat?
2. done
1.   ‘What did you do?’

Among the varieties of Eastern Lombard where both the sentence-initial and the postverbal distribution are accepted, it can be consistently observed that the former is systematically heavier than the latter. A possible correlation is worth exploring in further studies.²¹ In this sense, our proposal does not entail the presence of two sets of wh-phrases (weak and heavy) as suggested by morphological accounts for Italian (e.g., Bocci et al. 2022), which is not desirable neither on theoretical nor on empirical grounds (Bocci et al. 2024).

The strong tendency to pronounce lower copies as a consequence of the general stress pattern also provides a plausible explanation for the presence of a generally rich low periphery in Camuno, which seems to be a common property of NIDs (Ledgeway 2020).

Finally, the presence of a sentence-finality requirement in Bellunese (see Munaro 1999) corroborates the hypothesis of a role of this type of constraint in the distribution of wh-phrases as a possible explanation for similar patterns in typologically close varieties.