Embu is optional in Wh- questions (cf. (1a)–(1g) and (2)–(4)), except for Wh- arguments introduced by the dialectal Wh- word inda ‘what’ (cf. (1h) and (1i)), where it is obligatory in the contracted form ’mbu (GPT). Note further that embu/’mbu is not obligatory with other dialectal Wh- words/phrases in CG. It is also not obligatory with inda Wh- adjuncts (cf. (1f) and (1i)).

With complex inda Wh- phrases, ’mbu cannot occur between inda and the noun (5). As (4) (repeated below as (6)) shows, embu, but not ’mbu, can optionally occur in a position following the complex Wh- phrase.

The embu-strategy with the same asymmetric optionality characteristics is also used in embedded interrogatives. Consider the following:

Note that, not only the use of embu is the same in root and embedded questions, but, also, subject-verb inversion applies in both root and embedded embu Wh- questions (cf. (7) vs. (10) below).

The embu strategy also occurs in Yes/No questions. In these interrogatives, a non-Wh- element appears in the position preceding embu (cf. (11)).

Embu may not be used in Yes/No questions in which no such element precedes embu (cf. (13)).

The embu strategy, therefore, may be used to form a Yes/No question provided that an element appears in the clause-initial position preceding embu as in the case of Wh- words.

Embu may also appear in declaratives under certain conditions. There is, however, a disagreement in the literature regarding the status of the relevant data. On the one hand, Gryllia & Lekakou (2007), Fotiou (2009) and Agouraki (2010) have proposed that the embu strategy also occurs in focalizing constructions. GPT, on the other hand, take these cases to be rather marginal. They claim that (14) is ungrammatical, because movement of the focused clefted element is not allowed in CG clefts.

Although we tend to concur with GPT regarding the status of (14), similar sentences seem much improved, especially when a quantificational adverb precedes embu. (15) is such an example.

O Jannis is a topic in (15). In section 4.2.2, we will propose a unified analysis for these cases.

Although it is tempting to follow GPT, Panagidou (2009) and Agouraki (2010) in seeing embu as a contracted form of the copula en and the complementiser pu, there is a set of properties of this element that need to be looked at, before providing a syntactic analysis for it.

First, one would expect the putative copula in embu to inflect for tense, number and person as in standard copular sentences, contrary to fact (cf. Agouraki 2010; Kanikli 2011a; Papadopoulou 2014).

The alleged copula in embu does not inflect for past tense in (16). If, indeed, embu is the contracted form of the copula en and the complementiser pu, the copula appears only in the en form, which is the present tense form of both singular and plural third person in CG.

The same holds for the embu Yes/No question in (11), where embu does not agree for person and number with esi ‘you’ (cf. embu (11) instead of ise ‘are.2.SG’ pu ‘that’ in (18)).

Moreover, Kanikli’s (2011b) study showed that the inflection of en for tense results in highly degraded sentences (see Kanikli 2011b for a more detailed discussion).

It must be noted, though, that there is a discrepancy in the literature as to whether embu may or may not bear inflection. GPT and Panagidou (2009) argued that embu may inflect for tense, whereas we argue, on a par with Agouraki (2010), Kanikli (2011a) and Papadopoulou (2014), that embu may not inflect for tense (cf. (17)). The case is reminiscent of a similar disagreement, which exists in the literature of est-ce que interrogatives in Romance languages.

Obenauer (1977), Harris (1978), Lefebvre (1982), Blanche-Benveniste et al. (1984), Rooryck (1994), Cheng & Rooryck (2000) and others argue that est-ce que is inert for inflection, whereas others, such as Langacker (1965), maintain that est-ce que may bear inflection (cf. (22) and (28)). Munaro & Pollock (2005) claim that est-ce que is inert for inflection only in embedded interrogatives and que, pourquoi and comment root interrogatives.

As far as European Portuguese Wh- questions are concerned, Duarte (2000) argues that é que cannot bear inflection (cf. (19) and (21)), whereas Soares (2003) suggests that it can.

The following sections present the est-ce que Romance interrogatives and the different approaches to the status of est-ce que and é que in these structures.

As already argued, verb movement to C7 could not be triggered by the need to satisfy the Wh- Criterion. A relatively standard account runs as follows: verb movement to C is triggered under agreement by an EPP feature (Chomsky 2004; 2005). We assume that C bears an uT feature (Pesetsky & Torrego 2001) and an EPP feature in both root and embedded interrogatives in CG. T, agrees with C checking uT, and moves to C satisfying its EPP feature.

Wh- movement is also triggered under agreement with C by EPP on C. Chomsky (2000; 2001) proposes that the Wh- element bears an uWh feature and a Q(uestion) feature. C bears a Wh- feature and an uQ(uestion) feature. The Wh- element Agrees with C checking its uWh feature and the uQ feature on C. Assuming Wh- movement and I-to-C movement to be independently triggered offers an insight into languages where the fronted Wh- element may occur with an overt complementiser, but I-to-C movement is banned (see the nonstandard Italian and French varieties data discussed in Benincà 2001); languages where the fronted Wh- element may occur with an overt complementiser and I-to-C movement is obligatory (e.g. CG embu questions); languages where the fronted Wh- element may not occur with an overt complementiser and I-to-C movement is obligatory; and languages where the fronted Wh- element may not occur with an overt complementiser, and I-to-C movement is banned. This analysis also accounts for asymmetries between root and embedded interrogatives in the occurrence of I-to-C movement by attributing a different feature specification to root and embedded C.

Despite the advantages of this analysis, the trigger of Wh- movement and I-to-C movement as described above is problematic. Assuming a single C projection and assuming that an EPP feature is the only element capable of inducing movement, entails that the uT and the uWh feature on C are accompanied by an EPP feature, which triggers under agreement verb and Wh- movement respectively in CG. This could mean that either the features uT, uWh bear another feature; an option generally excluded by the theory, as it would amount to a reintroduction of the concept of feature strength; or that the EPP feature on C may be multiply satisfied, as in multiple subject constructions. This solution, though viable, is inelegant in that there is no way to specify exactly how many times the EPP feature on C could be satisfied. It would also require to independently rule out multiple Wh- movement in languages that have inversion, but are not multiple Wh- movement languages (say, English).

A more viable solution to this problem would be to assume that CP does not consist of a single C head, but an array of functional heads which may bear EPP features capable of inducing movement of heads or maximal elements to the CP. On this assumption, the EPP feature which triggers Wh- movement and the one which triggers subject inversion are carried by two different C projections. This approach can be formally captured in terms of Rizzi’s (1997) theory of a Split-CP. The following section defends a corresponding analysis for the CG data.

Consider the example in (41).

The constituent to vazo ‘the vase’ in (41) can receive neither a contrastive nor an information focus interpretation. In fact, the movement of this element to a clause initial position leads to ungrammaticality, an observation also pointed out by GPT, Fotiou (2009) and Agouraki (2010).8

In CG an element is eligible to be interpreted as information focus in its base position. Consider the following example.

The structure in (42) can be an answer to the question Ti espases? ‘What did you break?’. Nevertheless, it would be infelicitous if it was uttered as expressing identificational focus (cf. Kiss 1998). In CG an element may receive an identificational focus interpretation only by being hosted in a cleft structure (cf. Kanikli 2016). Consider the example in (43).

Although a DP cannot undergo movement to a clause fronting position in CG, quantifiers can (cf. Agouraki 2010). Compare the example in (41) with the example in (44).

Note that the only non-Wh- element that we found to occur in the pre-embu position was the quantifying adverb panda ‘always’ (15).

It seems that the elements which may undergo movement to a clause fronting position in non-embu structures, may do so in embu structures as well, qualifying for the pre-embu position. Wh- elements and quantifiers may move to a clause fronting position in non-embu structures, and they may do so in embu structures as well. DPs may not (41), so they may not do so in embu structures either (14). The question which arises is why DPs cannot move to a clause fronting position in CG.

One could assume that DPs cannot move to a clause initial position in CG, because they do not bear a focus feature. As already noted, the Focus projection in the CP is activated only when there is an element bearing a Foc feature (Rizzi 1997). The proposal, however, that DPs do not undergo focus movement in CG is contradicted by the fact that they actually do in clefts. Focalized DPs in clefts display binding effects, which suggest that they are not base-generated in the cleft clause, but move there from their base position in the embedded clause. Consider the example in (45).

The clefted constituent in (45) is a PP which involves the pronoun tis ‘her’. The pronoun is bound by the DP i Maria. This suggests that the clefted constituent originates inside the pu clause, where it can be bound by the subject i Maria. DPs, therefore, undergo focus movement in CG, but they may do so only in clefts.

In view of these facts, we cannot assume that embu is a realization of a Foc head in Rizzi’s (1997) array of C projections. If this was the case, DPs should be able to move to the pre-embu position.

The idea that Wh- elements are quantifiers is certainly not a novel one. Chomsky (1977) and a wealth of subsequent syntactic work have shown that they behave like operators. The semantics of questions based on Karttunen’s (1977) seminal work also analyses Wh- elements as existential quantifiers. Given the fact that only Wh- elements and some quantifiers may move to a clause initial position in CG, we propose that the C head lexicalised by embu bears an uninterpretable quantification feature9 (uQu) and an EPP feature. This allows Wh- elements, which bear quantification (Qu) features as part of their lexical specification (Dobrovie-Sorin 1990), to agree with the C head and move to its specifier to satisfy the EPP.

Under this analysis, DPs cannot undergo raising to a clause fronting position in CG, because they are not lexically defined as quantificational. Nevertheless, they may undergo focus movement in clefts, as the structure where a DP can be structurally defined as a syntactic quantifier is provided (Dobrovie-Sorin 1990).10

Apart from an uQu and an EPP feature, we propose that the C head lexicalised by embu bears an interpretable interrogative feature: Q. The Wh- element bears, apart from a Qu feature, an uninterpretable interrogative feature in Wh- questions, uQ, which renders it active in the derivation. C Agrees with the Wh- element and the Wh- element moves under agreement to its specifier position in order to satisfy its EPP feature. Given Chomsky’s (2000; 2001) activity condition, we predict that in an interrogative structure like (46), in which a quantifier and a Wh- element co-occur, the Wh- element, which bears uQ, will move, under agreement with the relevant C projection.

The above analysis further accommodates the fact that quantifiers stay in situ in some structures, whereas in others they move to clause initial positions in CG. The analysis predicts that in the latter case, the quantifiers bear ufeatures related to discourse related interpretation such as uEmphasis (cf. Agouraki 2010).11

Having accounted for the properties of the CG data with respect to which elements are eligible to move to a clause initial position, and having stated which features we assume to be involved in Wh- movement in (embu) Wh- questions in CG, the following section proceeds to provide a structural analysis for the embu questions.

Consider the example in (47) and the associated structure in (48).

The subject o Jannis in (47) occupies Spec,vP as V-to-I movement satisfied the EPP on I⁰ (Alexiadou & Anagnostopoulou 1998). Apart from the subject, all the other elements in (47) are in the CP domain. This accounts for the fact that sentential adjuncts may only follow the verb in embu Wh- questions (see section 3.1).

We propose that the verb occupies the lowest C projection in Rizzi’s (1997) array of functional projections: Fin⁰ (see (40)). We consider that Fin⁰ bears an uT feature (Pesetsky & Torrego 2001) and an EPP feature. I⁰, where the verb is hosted, bears an interpretable T feature. I⁰ Agrees with Fin⁰ valuing its uT feature and, under agreement, I⁰ (and concomitantly the verb) moves to Fin⁰ to satisfy its EPP feature.

As already proposed, embu is a single C head. We argue against an analysis of embu as consisting of two C heads (cf. the analysis Soares 2003 proposed for the European Portuguese é que). Such an analysis is undermined by the fact that nothing may intervene between en and pu (assumed to be contracted to embu). Moreover, the fact that nothing may intervene between the Wh- element and embu suggests that they “doubly-fill” the same projection. We take embu to be base generated in the head position of a Wh- projection and the Wh- element pcon to move to its specifier. As previously argued, the Wh- projection bears an interrogative feature (the Q feature), an uninterpretable quantification feature (the uQu feature) and an EPP feature. The Wh- element pcon bears an uninterpretable interrogative feature (uQ) and an interpretable quantification feature (Qu). Pcon in (47) Agrees with the Wh- projection and under agreement moves to its specifier position to satisfy its EPP feature.

In terms of Rizzi’s (1997) array of functional projections (40), the Wh- projection occupies a position higher than Fin⁰ and lower than Force⁰.12

According to the analysis defended above, embu is an overt Wh- head. The difference, therefore, between the two Wh- variants in CG boils down to whether the Wh- head is null or overt respectively. This accounts for the fact that embu and non-embu Wh- structures share the same syntactic properties. As shown, they both involve subject inversion. They interact with other left periphery elements in the same way and the distribution of adverbs is the same in both Wh- structures.

On the assumption that embu is an overt Wh- head, what still remains obscure is why ’mbu occurs obligatorily in inda Wh- arguments. In other words, why a C head must be obligatorily realized in a certain interrogative, whereas in others its overt realization is optional. The next section addresses the asymmetry in the obligatory occurrence of ’mbu and proposes an account for it in the light of the syntactic behavior of inda Wh- arguments in sluices.

When we consider sluicing, we observe that while embu is banned in sluices (cf. (50a) and (53a), ’mbu occurs obligatorily along with inda as a sluicing remnant (cf. (52a) and (54a)).

The above examples are sluices in matrix ((50) and (52)) and embedded clauses ((53)–(54)) (Merchant 2001).13 Following a deletion approach to sluicing, sluices involve Wh- interrogatives where a part of the structure is deleted. In the examples in (50) and (53), embu cannot be pronounced. In other words, it belongs to the part of the structure which must be deleted. On the contrary, ’mbu in the sluiced inda Wh- arguments in (52) and (54) cannot be deleted. Its deletion results in ungrammaticality as the examples in (52b) and (54b) show. Therefore, ’mbu is not part of the structure which undergoes deletion in sluices.

In our analysis, embu is a C head, namely, Wh⁰. The data in (50) and (53), therefore, suggest that sluices are Wh- interrogatives where only the element occupying the Spec,CP (namely the Spec,WhP) is allowed to be a sluicing remnant. The rest of the structure must be deleted (see the ungrammaticality in (50a) and (53a)). This fact, however, is at odds with deletion approach assumptions regarding sluicing. According to Merchant (2001), sluices are Wh- structures where IP is deleted. The data in (50a) and (53a), though, suggest that apart from IP, the complementiser embu must be deleted.

The above data is not the only problematic data for an approach to sluicing as IP deletion. There is cross-linguistic evidence that elements which move to C (e.g. I-to-C movement), as well as base-generated complementisers are not allowed to be pronounced in sluices (cf. li, the overt interrogative C in Serbo-Croatian, Boškovic 1997, which is also illegitimate as a sluicing remnant, Ochi & Hsin 1999). In order to account for these data Merchant (2001: 62) put forward the Sluicing-COMP generalization, according to which, ‘no non-operator material may appear in COMP’ in sluicing.14 The generalization captures the data in (50) and (53), where only the Wh- element is eligible to occur as a sluicing remnant.

Granted that only the element occupying Spec,WhP may occur as a sluicing remnant in Wh- sluices, the fact that ’mbu can occur along with inda in the Wh- sluices in (52a) and (54a) undermines the idea that ’mbu is an allomorph of embu, and suggests that inda and ’mbu are in fact a single Wh- word: indambu. On this assumption, the obligatory occurrence of inda with ’mbu in Wh- arguments is accounted for. Inda and ’mbu were reanalysed into a single Wh- word, indambu, that is why inda cannot occur without ’mbu neither in Wh- arguments nor in Wh- sluices. Therefore, there is no syntactic asymmetry between CG Wh- questions with respect to the occurrence of embu (Wh⁰), simply because ’mbu in (i)ndambu (and the other contracted forms nambu, ambu) is not a Wh- head.

Having proposed an analysis for the embu Wh- questions, the following section shows how the analysis extends to French and European Portuguese interrogatives.