A Special (Question) View on Wh-Doubling in Lombard Varieties

: In this article, we exam a speciﬁc type of Wh-Doubling (WhD) that can be observed in some Northern Lombard varieties. Differently from all the previous types of WhD described for Lombard and other Northern Italo-Romance varieties, in the phenomenon we analyze, the two wh-forms are identical and display the - E morpheme. Furthermore, in all the varieties where this type of WhD is present, it is associated with the encoding of special questions and cannot be used in true requests for new information. We propose an analysis of this construction assuming a split-CP representation where the higher wh-form activates a functional projection involved in the expression of special questions. We also hypothesize that this construction derives from an interrogative cleft; if this hypothesis is correct, the formation of the - E morpheme and the re-analysis of the cleft as monoclausal are the two sides of the same process.


Introduction
In this contribution, we analyze and discuss a special type of Wh-Doubling (WhD) that can be observed in some Northern Lombard varieties.We have two main goals: present novel data about the WhD phenomenon that have not been discussed before in the literature, and show that the cases we take into examination can shed light both on the origin of the allomorphy of wh-items in Lombard varieties and on the nature of the WhD itself, in particular about its relevance for the syntax/semantics interface in the encoding of non-standard questions.
The WhD is a peculiar phenomenon that is displayed across the Northern Italo-Romance domain with some variation (cf.(Munaro and Poletto 2023) for a recent overview).In general, it can be described as the presence of two separate wh-items in the same interrogative clause that introduce the same variable.One is usually on the left periphery of the clause; the other is in situ: As we will discuss more in detail in Section 3, a property that all cases of WhD described before share, is that the wh-item in the left periphery is phonologically and morphologically 'lighter' than the one found in situ.In (1), for example, sa is a monosyllabic clitic form, while cossa is a full bisyllabic wh-form.However, some Northern Lombard varieties, where the WhD is very widespread, can also have questions like (2), where in both positions there is a full form like ku'zE 'what': ( All varieties that have structures like the one in (2) share a relevant property: these questions are not standard root interrogatives, asked in order to assign a value to the variable, but are special interrogative clauses that have a more specific pragmatic interpretation, which we will define on the basis of Obenauer's (2004) typology of non-standard questions.
The paper has the following structure: in Section 2, we briefly indicate our sources; in Section 3, we provide a general overview of the WhD phenomenon; in Section 4, we describe a specific type of WhD, which we label 'Strong Wh-Doubling'; and in Section 5, we present a formal analysis of the structure of Strong WhD.

Materials and Methods
All the data we discuss in this paper are taken from Bernasconi (2021), who has collected (through interviews in presence) examples of questions from Lombard varieties in the area around Como in Italy, and in the Ticino in Switzerland.The collected sentences are translations of stimuli presented in Italian (the questionnaire contained 60 stimuli).When necessary, Bernasconi has also discussed interpretative nuances and alternative translations with the speakers in order to clarify whether different constructions are optionally interchangeable or whether there are specific semantic differences.Bernasconi's work has taken into account the varieties spoken in the following points: Davesco (near Lugano, CH), Mendrisio (CH), Olgiate Comasco (in the province of Como), Cuasso al Monte (Varese), Solbiate con Cagno (Como), Uggiate-Trevano (Como), Ronago (Como).

Preliminary Facts about Wh-Doubling
The Wh-Doubling (WhD) is a phenomenon observable in some Northern Italo-Romance varieties.In this section, we will describe its properties and the variation it displays across the Northern Italo-Romance domain.
A question has WhD when it contains two (usually morphologically different) whitems that introduce the same variable, so it is not comparable with the cases of multiple wh-items, as those found, for example, in Slavic languages (Rudin 1988), where each whitem refers to a different variable.The phenomenon seems to be typologically rare, but there are similar constructions in adjacent German varieties, like the Swiss German dialect of Uri (Frey 2006) In the dialect of Uri, the WhD is completely optional, even if it is the preferred variant to express a true information-seeking question.Notice that in this case the obligatory wh-item is found on the left periphery of the clause.
In the Northern Italo-Romance domain, it is possible to find structures like those exemplified in (4): These examples from a Lombard variety spoken in the province of Lecco show the main properties of the WhD.The two separate wh-items are morphologically different; one is on the left periphery of the clause, while the other is in situ.We start our description with this last property.
The term in situ is usually applied to wh-items when they surface in the syntactically unmarked position of the corresponding argument or adjunct.It should be pointed out, however, that in the vast majority of cases of WhD reported in the literature, the second wh-item is the final element of the interrogative clause, like in (4).There are some exceptions, like those reported in ( 5), but usually it is not clear from the data description if the material following the wh-item in situ is in its unmarked position or if it is right dislocated/extraposed.
( This point is potentially very relevant for a formal syntactic analysis of the WhD (cf.Poletto andPollock 2004, 2009), but it is not crucial for the specific type of WhD we examine in this contribution.We will come back to this aspect in Section 5.
The second relevant property of the WhD is that the two wh-forms are generally different.In order to describe this characteristic in a proper way, some preliminary facts have to be taken into consideration.Italo-Romance varieties very often present different allomorphes for the same wh-item.Even standard Italian has separate forms for the whitem corresponding to 'what': che, cosa, and the more formal che cosa.In the Lombard varieties we are taking into consideration, some wh-items have two or even three different forms.In general, this type of allomorphy is found in the wh-items corresponding to 'what', 'where', and 'how', but there is much micro-variation regarding the exact items and the number of corresponding allomorphes.As an example, we can take into consideration the variety of Mendrisio (a Lombard dialect spoken in the Ticino area in Switzerland; cf.Poletto andPollock 2004, 2009;cf. also (Bernasconi 2021), who has found that some speakers do not have the form me for 'how').Here, the wh-items corresponding to 'what', 'where ', and 'how' have  Paradigms like the one in (6) have been analyzed by Poletto and Pollock (2009) in terms of the tripartition of pronouns in clitic/weak/strong forms proposed by Cardinaletti and Starke (1999).Thus, sa/ndu/me are clitics, kuza/indua/kuma are weak pronouns, while ku'zE/indu'E/ku'mE are strong pronouns.The main difference between the forms in (6b) and (6c) is that the former have a restricted syntactic distribution since they have to appear adjacently to the verb and cannot be used in isolation (but see also Manzini and Savoia (2011) on this).On the other hand, they cannot be considered clitics like the forms in (6a), since they are bi-syllabic and carry stress (cf.Munaro and Poletto 2023).There is variation across the Lombard domain regarding the paradigm of wh-items.For instance, the variety spoken in Cuasso al Monte, in the province of Varese, has a clitic form only for 'what', while 'where' and 'how' display only two forms, as shown in ( 7 There is an interesting implicational order of the wh-items that present allomorphy in the sense that if a variety has multiple forms for one wh-item, it is the one corresponding to 'what', if it has multiple forms for two wh-items, the second one is generally 'where', and if it has three wh-items with allomorphy, the third one is 'how' (cf.Benincà and Poletto 2005 on this).The other wh-items do not have this property and are excluded from WhD constructions.One possible exception found by Bernasconi (2021)  Independently from the exact paradigm of wh-items, WhD constructions usually display two different allomorphs.More precisely, the wh-item in the left periphery is the "weaker" form in Poletto andPollock's (2004, 2009) terms, in the sense that it is either a clitic or a weak pronoun like those in (6b); on the other hand, the in situ form is the strong one, like those in (6c).This distribution is observable, for instance, in examples (4a-b), where there is no particular interpretative difference between the two variants.Manzini and Savoia (2011), who do not accept an analysis in terms of a clitic/weak/strong divide of the paradigm, report some cases of in situ wh-items that would be weak according to the divide, and also cases of WhD with identical forms: However, all the previous studies clearly state that WhD with two forms, like ku'zE 'what', or with a "weaker" form in situ and a fronted "strong" form, is impossible in all the described varieties of Lombard.Since we present here cases of this type, it is necessary to add some information about these '-E series' forms.As mentioned above, Poletto andPollock (2004, 2009) argue that wh-items of the '-E series' are strong pronouns, while Manzini and Savoia (2011) reject this view and consider them standard wh-items carrying a special -ε morpheme associated with Focus; according to this analysis, the [Focus] feature provided by -E is equivalent to the one associated with wh-movement to the CP.We will present our view on the '-E series' in the analysis section, but here it is important to point out that the -E morpheme derives from a grammaticalized copula, so that in origin ku'zE would have been segmented as kuz(a) ε 'what is' (this has been proposed for the first time by Munaro 1999, pp. 204, 209, 226-27).A strong piece of evidence for this diachronic origin is provided by Donzelli and Pescarini (2018), who discuss the WhD in two Lombard varieties, and among these is the one spoken in Cavergno, in the Valle Maggia in Switzerland.The dialect of Cavergno has a wh-item for 'what', which appears to be derived from a chunk of an interrogative cleft, kuz-E-u, with both the copula E and the 3sg subject enclitic u.This form appears in situ in questions like (10), which is a case of wh-item in situ without WhD: According to the analysis we propose in this contribution, the data we provide constitute further evidence for the cleft origin, as discussed by Donzelli and Pescarini (2018).
There are two other phenomena connected to the WhD.The first is the construction labeled 'Operator Wh-Doubling' by Munaro and Poletto (2023).In this special type of WhD, the wh-item in the left periphery is a form of 'what', while the one in situ is the questioned one.This structure is found not only in Lombard but also in Venetan varieties, where it is usually associated with special questions, whose pragmatic function is not to assign a value to the variable but to express disapproval for the interlocutor's actions (cf.also Obenauer 2006) We will not examine this type of WhD since it is a different phenomenon from the one we concentrate on (cf.Munaro and Poletto 2023); it lacks a complementizer-like element after the higher wh-item, and it follows the general rule of the two different forms.However, it is noteworthy that in Venetan, it usually has a similar interpretation to the WhD cases we examine in Lombard.
The second phenomenon that has always been considered alongside the WhD in Northern Italo-Romance varieties is the wh in situ construction, without a second wh-item in the higher position.The WhD and wh in situ have been linked and considered surface variants of the same syntactic phenomenon at least since Munaro's (1999) work on the distribution of wh-items in Bellunese and other Northern Italo-Romance dialects (see also Munaro and Poletto (2023) and Manzini and Savoia (2011) for a different point of view).As we will show in the next section, the WhD we analyze does not have an in situ variant with a null or silent wh-item in the left periphery, since its peculiar properties are present only if the higher wh-item is spelled out and if it displays the -E morpheme.This, in turn, supports the idea that there are different types of both WhD and wh in situ, which should be analyzed in a separate way independently from the superficial similarities.
Finally, recent works on the WhD have shown that it is not a root phenomenon, as it can be found also in questions embedded under verbs like 'ask' or 'not know' (Manzini and Savoia (2011)), and it can also appear with long wh-movement (many cases have been described by (Bernasconi 2021) There is variation among varieties and, in some cases, even among speakers of the same dialect in relation to the acceptability of similar examples.For the goals of our contribution, it is important to stress that in all these cases the distribution of the two different forms is always the same: the higher form is the short or light one, while the form with the -E morpheme is found in situ.

The Strong Wh-Doubling
In this section, we describe the syntactic and semantic properties of a specific type of WhD, which we label 'Strong Wh-Doubling', which is characterized by the presence of two separate wh-forms displaying the -E morpheme.

Syntactic Properties
The main property of the strong WhD is that both the questioned wh-items are represented by the allomorphic forms with the final stressed -E.As we have already stressed, previous works on the WhD have considered this combination impossible in all the Lombard varieties taken into consideration.Only two examples are reported by Munaro (1999, pp. 210, 225) for Mendrisio, who considers them cases of incipient monoclausal structures derived from the corresponding biclausal ones.However, Bernasconi (2021)  This type of WhD can also be found in long wh-movement contexts, like in example ( 14), while it is excluded from embedded questions.In other words, it is a root phenomenon, a fact we would link to its interpretative properties.In all the cases of Strong WhD, there is a complementizer ke/ka immediately following the higher wh-item.The presence of this element could be considered evidence that all these cases do not involve two separate forms of the -E series, but that we are dealing with interrogative clefts, where the higher wh-item displays the standard, or 'weak', form followed by the copula.There are, however, some reasons to reject this analysis.The first one is that Strong WhD is found only in root questions, while interrogative clefts are possible in embedded ones.In (15) Furthermore, in all the cases reported in (13), there is no subject clitic preceding or following the presumed copula.Recall that in the variety of Cavergno examined by Donzelli and Pescarini (2018), the subject enclitic has become part of the wh-item kuz-E-u 'what', derived from a reduced interrogative cleft.In our cases, however, there is no visible subject clitic independent of the behavior of the dialect under consideration regarding the position of subject clitics in questions.Notice also that Bernasconi (2021)  We believe that the opposition between ( 16), with the subject clitic, and (13a), where it is absent, shows that the latter has a different structure, and more precisely that it is mono-clausal with a wh-item bearing the -E morpheme in the higher position.
A final proof in favor of this solution is provided by the variety spoken in Davesco, Switzerland, where there is a complementizer-like element following the higher wh-item even if it lacks the -E morpheme.The analysis we propose in Section 5 will also take into consideration the role of this complementizer-like element.Before presenting our analysis, however, it is necessary to briefly discuss the interpretation of questions with the Strong WhD.

Interpretative Properties
All the examples of Strong WhD we have presented so far share a relevant interpretative property, i.e., they are not standard questions, asked with the goal of attributing a value to the variable bound by the wh-item.In order to characterize the semantic side of these clauses, we adopt the descriptive typology of special questions proposed by Obenauer (2004Obenauer ( , 2006)).In particular, the Lombard questions with Strong WhD belong to two of the types analyzed by Obenauer: 'Surprise-disapproval Questions' (SDQs) and 'Can't-find-thevalue-of-x Questions' (CfvQs).The special interpretation of these interrogative clauses is indicated by the final <?!> notation in all the examples.
Obenauer's typology is based on the observation that many languages present specific grammatical properties in questions whose primary function is not a request for information, i.e., the attribution of a value to a wh-item.The first system examined from this point of view is represented by some Northern Venetan varieties, where wh-items can surface in different positions (Munaro 1999 and subsequent work).More precisely, Munaro has shown that in varieties like Bellunese, bare wh-items appear in situ in standard questions: (18) à =tu incontrà ki? have.2SG=you met who 'Who did you meet?' (Belluno, Venetan, (Munaro 1999)) There are, however, cases of questions with bare wh-items where the latter appear in the initial position.Crucially, these constructions cannot be interpreted as standard questions, i.e., as requests for new information.Obenauer (2004Obenauer ( , 2006) ) has shown that Bellunese provides reasons to distinguish at least three types of special questions with bare wh-items in the initial position: (a) Surprise-Disapproval Questions (SDQs); (b) Rhetorical Questions (RQs); (c) Can't-find-the-value-of-x Questions (CfvQs).
SDQs can be described intuitively as obligatorily expressing the speaker's attitude towards the event, an attitude of surprise usually combined with disapproval.The value of the wh-variable is known to the speaker or is not relevant.RQs convey, rather than a request for the value(s) of a variable, an assertion of opposite polarity, i.e., that corresponding values do not exist.Finally, CfvQs express that the speaker is not able to find the value(s) of the variable bound by the wh-operator.The latter are more similar to standard questions, but in Bellunese they are marked by the wh-fronting like the other two types.
The peculiar syntax of special questions in Bellunese has been interpreted by Obenauer as evidence that in special questions the wh-item is not in the standard position for interrogative wh-items, but has moved to a higher specifier, in the discourse-oriented layer of a split CP.More precisely, Obenauer assumes the analysis of wh in situ proposed by Munaro et al. (2001), who argue that examples like (18) are derived through remnant movement of the TP to ForceP or another high projection in the CP (with the vP moved to a Topic position).According to this view, in examples like (19), the wh-item is even higher, which can be considered evidence that special questions are encoded by the activation of dedicated projections in the upper part of the left periphery of the clause.According to Obenauer, an example like (19) has the following structure, where SDP is the projection specifically dedicated to the interpretative features of SDQs (labels are modified according to current cartographic representations; T-to-Force is represented without intermediate copies): The idea that interpretative features that are relevant for the pragmatic component are directly encoded in syntax is a crucial ingredient of this type of analysis.Languages like Bellunese activate these features through wh-movement, but there are other means, like, for instance, clause particles (i.e., using Merge instead of Move).Bianchi and Cruschina (2022), who adopt a different type of analysis based on the syntactic formalization of the Speech Act by Speas and Tenny (2003), offer a recent overview of special question particles in the Italo-Romance domain.For our analysis in the next section, we broadly follow Obenauer, but our main goal is not the semantic/pragmatic side of special questions, but rather the role of these constructions in the type of WhD we take into exam and in the formation of -ε series of wh-items in Northern Lombard.

A syntactic Analysis of the Strong Wh-Doubling
For our analysis, we need to recall all the elements we have described in previous sections.The Strong WhD has the following main properties: (a) From the syntactic point of view, it is characterized by the presence of two wh-forms of the -E series (or two standard forms in the variety spoken in Davesco), one in initial position immediately followed by a complementizer-like element, the second in final position; (b) From a semantic point of view, all the cases of questions with Strong WhD described by Bernasconi (2021) are either SDQs or CfvQs, assuming the typology of special questions developed by Obenauer (2004 and subsequent works).
Given that in the Strong WhD the two forms are identical, we exclude an analysis in terms of a 'Big DP', as in Poletto and Pollock (2009), with the two forms generated together and then moved to different positions in the clause structure.Following the reasoning by Manzini and Savoia (2011) for the other types of WhD, we propose that in the strong WhD, there are two separate lexicalizations of the wh-item.However, differently from the other types of WhD, the higher wh-item also presents the -E morphology.In our view, this is an indication that the -E morpheme can encode different features, not only [Focus] as proposed by Manzini and Savoia, i.e., that it is ambiguous.In particular, the interpretation of questions displaying Strong WhD suggests that the higher wh-item encodes an interpretative feature connected to special questions, which we label [SpeakEval], by which we mean that it adds an evaluation from the point of view of the speaker to the interpretation of the interrogative clause.In a cartographical structure of the left periphery of the clause, the projection introducing [SpeakEval] is higher than ForceP in the CP sublayer containing discourse and speech act elements.A similar projection, Eval(uative)-Speaker, has been proposed by Hinterhölzl and Munaro (2015), who have compared the position of modal particles in German and Bellunese and have proposed that the point of view of the speaker is encoded immediately above ForceP.A question like (13b), repeated here as ( 21) would have the structure in ( 22 In (22) we do not examine in detail the internal structure of ForceP.Notice that in our analysis, the complementizer-like element ka heads the projection whose specifier hosts the higher wh-item.In our view, this 'doubly-filled Comp' configuration of SpeakEvalP is a consequence of a re-analysis process that has produced both the monoclausal nature of questions with Strong WhD and the -E series of wh-forms.In other words, questions with Strong WhD derive from true interrogative clefts that have been re-analyzed as monoclausal: the initial wh-item, combined with the copula, has undergone upward re-analysis from specifier of WhP (or FocusP, depending on the adopted structure of the split CP) to specifier of SpeakEvalP.The process is represented in ( 23 Notice the following aspects of our proposal: (a) In this representation, the process is triggered by the univerbation of the wh-item (moved from the embedded clause through the Focus position at the vP edge to the left periphery in (23a)) with the copula, i.e., by the formation of the -E morpheme.In the initial stage (23b) we can assume that -E encodes [Focus], as expected given that clefts are a type of focalization.In the following stage (23c), however, the morpheme is associated with [SpeakEval], a classic case of upward re-analysis as discussed by Roberts and Roussou (2003).We do not have diachronic data for these varieties, so it is not possible to have a confirmation about the trigger, but cases like those observed by Donzelli and Pescarini (2018) clearly indicate that the -E morpheme was formed from structures with a subject enclitic, i.e., with the copula moved above TP.(b) In (23c), we have inserted the second wh-item in situ at this point.Even if this is not crucial for our proposal, we can hypothesize that the WhD with a form from the '-E series' in situ is linked to the ambiguity of the -E morpheme, since the higher form does not encode the [Focus] feature anymore, but the one required by the special question interpretation.(c) As said above, we are not examining here the internal structure of the clause below [SpeakEvalP], but our proposal is compatible with both a true in situ position of the lower wh-item (like in Manzini and Savoia (2011)) and a remnant movement of TP analysis, like in the model developed by Munaro et al. (2001).However, a relevant point for a general theory of WhD and its relation with wh in situ is that if our hypothesis is correct, the in situ position has become available to wh-items following the formation of the -E morpheme, which allows a split between the scope position of wh-items in the left periphery and the [Focus] feature (Manzini and Savoia (2011)).
A further element in favor of the representation in ( 22) is provided by the distribution of the particle ma, likely borrowed from Italian.As pointed out by Bernasconi (2021), Northern Lombard varieties can encode special questions like SDQs and CfvQs also by the means of the initial particle ma, identical to the adversative particle, which has a similar function in Italian (cf.Giorgi 2018).However, ma and the Strong WhD are mutually exclusive: *ma ku'zE ka te= se drè a fa ku'zE?! Thus, (24a), where the WhD presents the standard configuration with the clitic form in the left periphery, not followed by the complementizer, is allowed, while (24b), with both the particle and the Strong WhD, is ungrammatical.This decisively supports the idea that ma and an initial wh-form with the -E morpheme have the same function and are merged in the same structural position (adopting Obenauer's (2004) terminology, they are alternative checkers): (25) [ SpeakEvalP ma [ ForceP se=te=se drè a fa ku'zE]] Finally, we have to discuss the examples found in the variety of Davesco, like (17a-b), where there are two wh-forms without the -E morpheme plus the complementizer after the higher wh-item.In our view, these cases represent a further step in the process represented in ( 23), where the [SpeakEval] feature is not encoded by the -E morpheme but simply by the direct merge of a wh-item in the specifier of SpeakEvalP.So, while in varieties like Bellunese, the projection where special questions are encoded is activated by moving the wh-item from a lower position, in the variety of Davesco, there is merge of a dedicated wh-item, similarly to what happens in other types of WhD.

Concluding Remarks
In this paper, we have described a specific type of Wh-Doubling found in some Northern Lombard varieties, which we have labeled 'Strong WhD'.The existence of a similar construction shows that Munaro and Poletto (2023) are right when they say that there are different types of WhD (and wh in situ), which do not necessarily need a unified analysis.In particular, we have tried to show that the Strong WhD is strictly connected to the expression of special questions and thus likely involves the activation of the higher part of the left periphery of the clause.

three different forms:
): is the variety spoken in Olgiate Comasco, where 'who' also seems to have two forms: . ): has found examples like the following in three of the Lombard varieties she has investigated: , this is shown both for standard Italian and a Lombard dialect: has described examples of true interrogative clefts in these varieties where the clitic is present: ): SpeakEvalP ku'zE [ SpeakEval ka [ ForceP te=dì ku'zE]]]