2.1. X-bar theory

In the 1980s (Chomsky Reference Chomsky1986, for instance), it was assumed that the clause was articulated around three heads: the verb (V), the inflection (I) and the complementizer (C) – the formulation of the X-bar format led to the structural hierarchy in (2):Footnote ³

Verb Second languages display a well-known root restriction according to which the finite verb occupies a second position and is preceded by exactly one constituent, be this the subject (3a), a complement (3b) or an adjunct (3c). As shown in StD (3), while all major constituents of the clause can occupy an initial position, only one such constituent can precede the finite verb: any combination of more than one constituent is ungrammatical (3d–f).

In a seminal analysis of these and other data, den Besten (Reference den Besten and Werner Abraham1983) exploits the fact that the CP layer in (2) is limited to exactly two positions, the specifier position of CP and the head C, to derive the V2 pattern elegantly. The first constituent of the V2 pattern is the specifier of CP, the finite verb is at C, and this configuration exhausts the CP layer.

In the literature it is generally acknowledged that even bona fide V2 languages display ‘exceptional’ patterns in which the finite verb is preceded by more than one constituent.Footnote ⁴ We refer to Broekhuis and Corver’s (Reference Broekhuis and Corver2016) detailed overview for discussion and additional references. As such, these patterns might be invoked as evidence for a more fully articulated LP, at least if all three elements preceding the subject position are contained in the LP: they could then be taken as evidence that the LP must contain more than just one specifier and one head. However, one common stance in the literature has been that in these apparent ‘exceptions to V2’ the (additional) initial constituent is analyzed as an extra-sentential constituent (‘satellite’, cf. Koster Reference Koster and S. J. Keyser1978). For Dutch, illustrations of these patterns concern, among other things, Dutch C(ontrastive) L(eft) D(islocation) with resumption by a demonstrative, as in (4a,b), speech act modifiers (4c,d), and irrelevance conditionals (d’Avis Reference D’Avis, Lohnstein and Trissler2004) (4e). If one adopts this approach, then these V3 patterns do not bear on the issue of the articulation of the LP, as they remain compatible with a parsimonious LP restricted to just Spec,CP and the C-head.

2.2. Cartography and the articulated LP

In the cartographic programme, initiated by Rizzi (Reference Rizzi and Haegeman1997) and Cinque (Reference Cinque1999), clause structure is decomposed into fine-grained articulations reflecting the semantic and/or pragmatic contribution of the structural layers, and, among other things, the unitary CP is replaced by what has come to be known as the articulated LP, where the Force layer which closes off the projection encodes the illocutionary force of the utterance and the Fin layer encodes the finiteness properties of the clause as replicated in the inflectional morphology on the verb. Additional projections encode discourse-related features such as [topic] and [focus]. Example (5), based on Rizzi (Reference Rizzi and Haegeman1997), is but one variant of a range of proposals for the decomposition of the LP.

Underlying ‘maximalist’ proposals for the richly articulated structures is the one feature-one head hypothesis, which maximalizes structure by postulating that each grammatically encoded interpretive feature is encoded on a distinct functional head.

With the advent of the articulated LP, the V2 phenomenon becomes of renewed interest. The minimal CP in (2) captured the V2 pattern elegantly, because the system provided just two positions: a unique specifier and a single head in the C-domain. On the other hand, the articulated CP – as in its implementation in (5) – potentially allows for five heads and related specifiers (or more, given recent proposals (see Rizzi and Bocci (Reference Rizzi, Bocci, Everaert and van Riemsdijk2017) and the references cited there)) and hence does not naturally account for the two constituent restrictions in the LP in V2 languages.

2.2. Proposals

There have been many attempts to reconcile the richly articulated LP structures with the structural restrictions regulating the V2 pattern. One such proposal, originally due to Haegeman (Reference Haegeman1996), is sometimes referred to as the ‘bottleneck’ effect. This author proposed that in root clauses the finite verb moves to the head Fin, and that by virtue of the V2 condition, the root Fin requires a specifier. Movement from V to Fin would be triggered by the finiteness features of Fin. In embedded clauses, these are reflected in the phenomenon of agreeing complementizers (see section 5.2.1 for a brief discussion). Locality restrictions mean that the filling of SpecFin blocks all additional LP movement, and FinP thus becomes a bottleneck. Haegeman’s proposal is adopted and adapted in Poletto (Reference Poletto2013) and Wolfe (Reference Wolfe2015a, Reference Wolfe2016). These authors develop a typology of V2 patterns based on the locus targeted by the finite verb and the accompanying initial constituent: Fin or Force. They distinguish Fin-V2 languages (6a) from Force-V2 languages (6b). In Force-V2 languages the finite verb moves to Force via Fin, and the initial constituent moves to SpecForceP via SpecFinP.

The predictions of the Wolfe/Poletto typology are that in Fin-V2 languages (6a), multiple access to the LP remains available, leading to attestations of V3 and V4 orders, as exemplified among others in Romance varieties (cf. Benincà and Poletto Reference Benincà, Poletto and Rizzi2004; Benincà Reference Benincà and Zanuttini2006, Reference Benincà, Bidese and Cognola2013 for illustration and discussion). In Force-V2 languages, the movement from SpecFinP to SpecForceP would be blocked by the additional LP movement of a constituent from within TP (Haegeman Reference Haegeman1996; Roberts Reference Roberts and Rizzi2004; Holmberg Reference Holmberg, Kiss and Alexiadou2015, Reference Holmberg, Theresa Biberauer, Wolfe and Woods2020). Directly merging an additional constituent in the CP area is also blocked, because such a constituent would hypothetically be inserted to satisfy a criterial feature and hence could not itself move to SpecForce. Once inserted, the constituent will then again block XP-movement from SpecFinP to SpecForceP. ‘Diverging’ V3 patterns will only arise when main clause-external constituents combine with ForceP (=(6c)) (cf. Broekhuis and Corver Reference Broekhuis and Corver2016 for ‘main clause-external’; Zwart Reference Zwart, M. den Dikken and Tortora2005 for ‘extra cyclic’; Astruc-Aguilera Reference Astruc-Aguilera2005 for ‘extra-sentential’; Haegeman and Greco Reference Haegeman and Greco2018 and Greco and Haegeman Reference Greco, Haegeman, Woods and Wolfe2020 for ‘frame setter’).

2.3. The LP and invariant die in the Ghent dialect

Obviously, the question arises how the Ghent example (1a) complies with the V2 constraint. Note that the sequence adjunct+die also requires that the finite verb precede the subject: (7a) with non-inverted subject verb order is ungrammatical. If we assume that (1a), repeated in (7b), satisfies the V2 constraint, the coordination data in (7c) and (7d) show that this is not because invariant die forms a constituent with the initial constituent.

Zwart (Reference Zwart1997) assimilates the Ghent die pattern to StD CLD (4a) and interprets it as a form of adverbial dan resumption (4b), with the initial constituent clause-external and die the first element in the V2 configuration, as schematized in (8a,b). For Zwart, then, invariant die in the Ghent variety has the same function as the resumptive adverb dan (8c).

While agreeing with Zwart’s analysis of dan resumption and the idea that the initial constituent preceding dan is an extrasentential constituent, we will argue that in patterns with invariant die ((1a) and (7b)), the initial constituent is clause-internal (9). The pattern will be shown to be compatible with the V2 constraint provided we adopt an articulated CP. Thus the data actually offer some support for assuming an articulated CP also in the Germanic V2 languages.

3.1. An extrasentential constituent

In the literature, resumptive root V3 patterns in which an adverbial adjunct precedes a demonstrative adverbial, which in turn precedes the finite verb, as illustrated in (10), are often viewed as the adverbial analogue of the better known CLD patterns (Zwart Reference Zwart1997, Hoekstra Reference Hoekstra1999, Broekhuis and Corver Reference Broekhuis and Corver2016).Footnote ⁵ Typically, a matching requirement determines the choice of adverbial: daar ‘there’ resumes locatives, dan ‘then’ and toen ‘then’ resume temporals. The patterns in (10) can then be seen as an instance of specialized resumption, in which the deictic adverb that resumes the initial constituent is specialized in that it matches the semantics of the initial constituent.

According to some prevalent analyses, the initial adjunct in these ‘V3 patterns’ is ‘extrasentential’, as shown schematically in (11). Crucially, this constituent is not merged clause-internally at any stage of the derivation and is inserted at the level of ‘discourse syntax’ (cf. Greco and Haegeman’s Reference Greco, Haegeman, Woods and Wolfe2020 FrameP). XP in (10) is a topical constituent, and as a consequence, the resumptive demonstrative pronouns (die, dat) and deictic adverbs (dan, daar, toen) are labelled variously topic pronouns (Hoekstra Reference Hoekstra1999), D-pronouns (Hoekstra Reference Hoekstra1999), or d-words (Koster Reference Koster and S. J. Keyser1978, Zwart Reference Zwart1997). The topical interpretation as well as the clause-external position of the initial constituent led to the correct prediction that neither negative constituents (12) nor wh-constituents (13) can function as first constituents in StD CLD, (12a) and (13a), or in adverbial (12b,c) and (13b,c) V3 resumption. This is so because (i) quantificational constituents are not suitable as topics (see Rizzi Reference Rizzi and Haegeman1997) and (ii) the clause-external position is not subject to reconstruction.

Obviously, there are no such restrictions for regular V2 in which the initial constituent is clause-internal and may be focal, as illustrated in (14), as well as topical:

3.2. Adverbial resumption: specialized resumptives vs. generalized resumptives

The resumptive patterns illustrated above illustrate what, following Meklenborg (Reference Meklenborg, Woods and Wolfe2020), we will call ‘specialized’ resumption. The main properties of adverbial specialized resumption are the following: (i) the specialized resumptive element (StD dan, toen, daar, etc.) is a regular adverb with its original adverbial meaning (i.e., it has the semantics of its non-resumptive adverbial use); (ii) the specialized resumptive adverb also occurs as an independent adverb in a V2 clause, both in initial (15) or mid-position (16); and (iii) the specialized resumptive adverb matches the initial constituent in terms of its semantic features.

What Meklenborg (Reference Meklenborg, Woods and Wolfe2020) labelled generalized resumptives are resumptive elements which are not subject to a matching requirement and are compatible with a wider range of initial constituents; the typical example is Mainland Scandinavian så ‘so’ (cf. Eide Reference Eide2011), illustrated by Norwegian (17):

At first sight, the generalized resumptive så in (17) might be thought of as a ‘bleached’ resumptive which takes over the function and position of all the specialized resumptives, as schematized in (18a), which would lead to the prediction that specialized resumptive da and generalized resumptive så are in complementary distribution, contrary to fact. As shown in (18b), these two resumptives may co-occur and they will then occupy different positions, with da obligatorily preceding så, given the ungrammaticality of (18c). The fact that the specialized resumptive da and the element så can co-occur raises the question to what extent both are ‘resumptives’: if da resumes the initial adverbial clause, then what does så itself ‘resume’? Though clearly Meklenborg’s (Reference Meklenborg, Woods and Wolfe2020) distinction between da and så is a valid one, we will refrain from adopting her term ‘generalized resumptive’ in our article to refer to elements, like die, that are compatible with a range of different initial constituents.

4.1. Invariant die is not a specialized resumptive

Ghent invariant die is not a specialized resumptive because it is compatible with a range of different adjuncts: temporal (19a), conditional (19b), locative (19c), goal (19d), causal (19e), result (19f), and rationale (19g). Analogues of (19) with die resumption are sharply ungrammatical in all other varieties of Dutch.

The pattern in Norwegian (18b) in which the specialized resumptive is followed by så is replicated in the Ghent dialect: a specialized resumptive (dan/daar) can co-occur with invariant die, showing that the Ghent dialect does not have the two elements in complementary distribution either, as shown in (20).

In the next section, we first look at specialized adverbial resumption patterns in the Ghent variety and then compare them with the patterns with invariant die.

4.2. Specialized adverbial resumption in the Ghent variety and invariant die

Recall that specialized resumptive adverbs are available as independent adverbs, and that the resumption pattern is subject to a strict matching requirement. As shown in (21), the specialized resumptives tons/dan/daar team up with a matching initial adjunct:

We will assume that the pattern in (21) is the adverbial analogue of StD CLD (in line with Koster Reference Koster and S. J. Keyser1978, Zwart Reference Zwart1997, Hoekstra Reference Hoekstra1999, and Broekhuis and Corver Reference Broekhuis and Corver2016). The initial dislocated XP is ‘clause-external’ and is the resumptive constituent that fulfils the first constituent requirement of a V2 language, as schematized in (22):Footnote ⁷

In contrast with the patterns in (21), invariant die is not subject to a matching requirement. In addition, invariant die is not used as independent adverb (23) (Vanacker Reference Vanacker and Joep Kruijsen1980). Our informants do not accept an antecedentless occurrence of invariant die. In all its attestations in our database, invariant die is immediately preceded by an adverbial constituent.Footnote ⁸

The constituent preceding invariant die may be one of the adverbs that also function as specialized resumptives. In all examples in (24), die is optional, and the specialized resumptive itself can function as an adverb:

4.2.1. The constituent left adjacent to invariant die is not a topic

In contrast with specialized (adverbial) resumptives (12), invariant die is compatible with a left-adjacent negative adjunct (25). Likewise, for some speakers, invariant die is compatible with a left-adjacent wh-constituent (26), again in contrast with the specialized resumptive (13):

The compatibility of invariant die with a quantificational constituent to its left and the fact that such quantificational constituents are not natural topics entails that the constituent left-adjacent to invariant die is not necessarily interpreted as a topic. This means, in turn, that invariant die is not (semantically and syntactically) equivalent to the specialized resumptives dan/tons/daar (pace Zwart Reference Zwart1997).

Pursuing this observation, the compatibility of negative and wh-adjuncts as initial constituents with invariant die leads us to the conclusion that the constituents left-adjacent to invariant die are not clause-external in the sense of Broekhuis and Corver (Reference Broekhuis and Corver2016). If they were clause-external, they would be unable to scope over the clause or ensure clause typing. Put differently, then, the pattern with invariant die is closer to ‘regular’ V2, which also allows for quantificational constituents in initial position (cf. StD (14)), and we propose that the constituent left-adjacent to invariant die occupies a left-peripheral specifier, a proposal we generalize to all occurrences of invariant die, including those with a non-quantificational left-adjacent constituent.

There are two pieces of additional evidence that the constituent left-adjacent to invariant die is not necessarily a topic. First, the constituent left-adjacent to invariant die can constitute an answer to a wh-question (27). This would be unexpected if this constituent were topical, that is, if it represented old or discourse given information: the answer to a wh-question typically constitutes new information.

Second, epistemic adverbs such as waarschijnlijk ‘probably’ or speaker oriented adverbials such as spytig genoeg ‘sadly’ do not constitute obvious topics and are incompatible with (specialized) dan resumption (28a), but are compatible with invariant die (28b) (cf. Broekhuis and Corver Reference Broekhuis and Corver2016).

These observations lead us to conclude that the pattern with invariant die differs interpretively and syntactically from that with the specialized adverbial resumptive. We postulate that while it may look like a V3 pattern, the die pattern can be assimilated to the regular V2 pattern.

4.2.2. The resumptive in adverbial CLD vs. invariant die

On the basis of the coordination patterns in (7), we concluded that invariant die does not form a constituent with the adjunct to its left. De Clercq and Haegeman (Reference De Clercq and Haegeman2018, Reference De Clercq, Haegeman, Fuzhen Si and Rizzi2021, Reference De Clercq, Haegeman, Clercq, Haegeman, Lohndal and Meklenborg2023) propose that invariant die is merged as a head in an LP position, as shown schematically in (29).

Specialized resumptives are compatible with focusing elements such as juist ‘precisely’, zelfs ‘even’ or alleen ‘only’, both in StD (30) and the Ghent variety (31). Specialized resumptives can also receive focal stress (LDG, p.c., 25.10.2023).

We conclude that the specialized resumptive has phrasal status. In contrast, invariant die is incompatible with focusing elements or with focal stress (LDG, p.c., 25.10.2023), in line with our proposal that die is a head:

Both in StD and in the Ghent variety, the specialized resumptive adverb moves to the LP when it can; if the LP slot of the root clause which combines with a dislocated adjunct is unavailable because another LP feature has independently been activated (e.g., wh-feature, imperative, etc.), the specialized resumptive adverb will remain in a middle field position, as in (33a) and (34a). These data suggest that the specialized resumptive is merged in the middle field (see Hoekstra Reference Hoekstra1999 for some arguments from StD).

In contrast with the specialized resumptive, invariant die is incompatible with a mid-position, even when the LP is independently activated, as shown in (35):

Table 1.

specialized resumptive (dan/tons/demonstrative pronoun) vs. invariant die

Table 1 Long description

Comparison of specialized resumptive forms, described as dan, tons, and a demonstrative pronoun, against an invariant form labeled die is intended, but the table contains no entries. Without rows, columns, or numeric values, no frequencies, proportions, or patterns can be identified. No group differences, trends, or rankings can be reported. Any interpretation about which form is more common or preferred is not possible from the provided material. Supplying the table body, including headings and cell values, is necessary to write meaningful.

Following the reasoning above, these data show that invariant die is not merged TP-internally; we propose that invariant die is a head merged directly in the LP.

Table 1 summarizes the main properties of specialized resumptives and of invariant die.

On the basis of the differences listed above, we conclude that the invariant die pattern in the Ghent variety must not be assimilated to specialized resumption (pace Zwart Reference Zwart1997). The two patterns are differentiated along the following lines:

1. (i) The specialized resumptive is a phrasal constituent merged in a TP-internal position and (being topical) is moved to the LP. It is the moved specialized resumptive that satisfies the V2 constraint; the constituent to the immediate left of the specialized resumptive is clause-external.

2. (ii) Invariant die is merged as an LP head; the constituent left-adjacent to invariant die is merged TP-internally and is moved to the LP. It is the constituent left-adjacent to invariant die that satisfies the V2 constraint and hence the constituent left-adjacent to invariant die is not clause-external.

The fact that a specialized resumptive and invariant die may co-occur is compatible with the proposed analysis. In (36) dan is a regular adverb filling the first position of the V2 pattern. It co-occurs with invariant die. Example (37) shows how deictic adverbs with a resumptive function remain compatible with invariant die.

In what follows we explore how our analysis can be handled in a cartographic approach to the LP.

5.1. The Wolfe-Poletto typology of V2 languages

Our hypothesis is that in terms of the Wolfe/Poletto typology of V2 languages (section 2.2), the Ghent variety of Dutch is a Force-Fin-V2 language: regular V2 root clauses are derived by V movement to Force via Fin, and by movement of a constituent to SpecForce via SpecFin (38):

Based on the evidence presented in relation to (i) scope and (ii) clause typing, we have concluded that the antecedent of invariant die must be clause-internal. Given its incompatibility with modification, we hypothesized that invariant die is not a phrasal operator, leading us to the hypothesis that die is inserted as an LP head in Force. In invariant die sentences, the finite verb itself obligatorily inverts with the subject (7a), leading us to the conclusion that it also occupies an LP head position, in other words, Fin:

The derivation in (39) can be summarized as follows: the lowest LP head Fin encodes the finiteness properties of the clause and attracts the finite verb (which will be left-adjacent to the subject position); the topmost LP head Force is occupied by invariant die. The obligatory presence of a constituent to the immediate left of invariant die, a “die second” constraint, follows from the V2 requirement associated with Force. In the dialect considered, V2 (V movement to LP) is restricted to root clauses; likewise, invariant die is a root phenomenon.

When invariant die resides in Force, the finite verb must move to Fin, as shown by the fact that it precedes the subject. This confirms that the Ghent variety is a Force/Fin-V2 language. We tentatively extend this analysis to other V2 patterns in Flemish, the only difference being that in the absence of die, the Force requirement is met by the finite verb.

In (37), invariant die co-occurs with a fronted specialized resumptive (section 4.1), which we assume to be phrasal. In (37d), repeated for convenience as (40), the specialized resumptive is locative daar; it is clause-internal. The PP in Sint Kruis, the ‘antecedent’ of the specialized resumptives itself is clause-external (Broekhuis and Corver Reference Broekhuis and Corver2016). Following Greco and Haegeman (Reference Greco, Haegeman, Woods and Wolfe2020), we assume that it occupies the specifier of the clause-external projection FrameP.

5.2. Integrating the analysis into a richly articulated LP

The preceding account develops the Wolfe/Poletto derivational typology of V2. As it stands, the account deploys only the projections ForceP and FinP, but ignores the one-feature-one head hypothesis underlying the cartographic enterprise in which discourse-related features such as [topic] and [focus] are associated with specialized heads and are at the basis of projections such as TopP and FocP (cf. Rizzi Reference Rizzi and Haegeman1997). At least in some cases, the constituent preceding invariant die is clearly associated with a specific discourse function, most notably perhaps in case of wh-items, such as those illustrated in (14c,d). Assuming the articulated CP, one may well wonder how – if at all – the additional projections in (5) come into play. Retaining our assumptions that die is located in Force and that the finite verb remains in Fin, our representation of die resumption with a wh-constituent left-adjacent to invariant die would be as in (41); this is not in line with the hypothesis that fronted wh-phrases, like focused constituents, are associated with the LP SpecFocP.

5.2.1. The enriched CP and criterial positions

Deploying the fully articulated LP entails enriching (41) with the relevant discourse-related projections, here FocP, i.e. a dedicated projection for focused constituents, which will host the wh-phrase. which will host the wh-f. If the V2 requirement in Germanic necessarily implicates both the heads Fin and Force, as argued above, the wh-constituent, which consists of a matching focus feature which moves to SpecFinP, and should then move further to SpecFocP. However, in order to satisfy the Force-V2 requirement, it should ultimately end up in SpecForceP, as shown in (42a), in which ‘wh-phrase’ stands for the relevant constituent. But (42a) is not a licit derivation. The feature [wh] is criterial in the sense of Rizzi (Reference Rizzi and Haegeman1997, Reference Rizzi2006) defined as in (42b): criterial features comprise, among others, [wh], [Top], [Foc], [Rel] and [Subj]. Crucially, the specifiers of criterial heads are halting positions: a constituent which has satisfied the wh-criterion by moving to SpecFocP is frozen in place (see also Hsu Reference Hsu2017).

The same problem arises if the constituent left-adjacent to die is a topical constituent, of course. The next section explores how this derivational paradox can be overcome.

Our proposal is inspired by Rizzi and Shlonsky’s (Reference Rizzi and Shlonsky2006, Reference Rizzi and Shlonsky2007) work on modes of satisfaction of the Subject criterion, which we briefly summarize before presenting our analysis. Rizzi and Shlonsky’s (Reference Rizzi and Shlonsky2006, Reference Rizzi and Shlonsky2007) proposal includes the hypothesis that the one-feature-one-head maxim inherent in the cartographic enterprise can be circumvented by virtue of a bypassing strategy and featural enrichment. For additional application of the same idea, see also Haegeman and Danckaert (Reference Haegeman and Danckaert2017).

5.2.2. SubjP, the Subject criterion and subject extraction

In line with the cartographic tradition, Rizzi and Shlonsky decompose the unique subject layer of the clause into two layers: TP and SubjP, (43). SubjP is a criterial projection: its head comes with a criterial feature, defined as in (42b) (Rizzi and Shlonsky Reference Rizzi and Shlonsky2006: 138). In (43b), we see how the Subject criterion is satisfied:

By virtue of the Subj-criterion, the position SpecSubjP has to be filled and its filler has to remain in place. This requirement blocks all subject extraction. This effect of the Subject criterion can be observed in French, in which extraction of a subject across the complementizer que is illicit (44a). Rizzi and Shlonsky (Reference Rizzi and Shlonsky2006, Reference Rizzi and Shlonsky2007) propose that the replacement of que by qui in (44b) is a reflex of a special mechanism for the satisfaction of the Subject criterion which allows the subject to skip SpecSubjP.

For Rizzi and Shlonsky (2007), qui in (44b) lexicalizes the lowest LP head Fin, which has been enriched with φ-features. Haegeman and Danckaert (2017) represent the enriched Fin as ‘Φin’. The head-to-head relation between Φin and Subj in (45a) is strictly local, and geometrically it is identical to that between SpecSubjP with Subj in (43b). By virtue of the local relation of Φin and Subj, the φ-features on Φin satisfy the Subject criterion, as shown in (45a). Rizzi and Shlonsky (2007: 138–139) restate the condition for criterial satisfaction (42b) in terms of a locality requirement as in (45b), by which a sufficiently local and enriched head can also satisfy the criterion.

Rizzi and Shlonsky postulate an additional requirement for those cases in which the criterion is satisfied by a local head, here Φin: the φ-features on Φin which serve to satisfy the Subject criterion have to be independently licensed. The additional licensing is achieved by the moved subject, here a wh-phrase: on its way to the LP landing site which hosts the criterial wh-feature, the subject moves through SpecΦin and licenses the φ-features of Φin. Spec Φin is not criterial, so the subject can move further.

5.2.3. Invariant die and enriched Force

Consider (46a), with the wh-phrase wanneer ‘when’ left-adjacent to invariant die. Let us assume that wanneer ‘when’ carries a criterial wh-feature that has to be associated with an LP criterial FocP. Let us also continue to follow Wolfe’s proposal (Reference Wolfe2015a, Reference Wolfe2015b, Reference Wolfe2016) that the V2 property of Force-V2 languages requires that Force be overtly spelt out and that it must have a specifier. As discussed, (46b) is not a licit derivation because once wanneer has satisfied the criterion in SpecFocP, it should be frozen for movement.

The problem for deriving (46a) is that the derivation should satisfy the wh-criterion in FocP while at the same time satisfying the Force-V2-condition by moving wanneer to SpecForceP. These two movements are incompatible because once in SpecFocP, wanneer satisfies the wh-criterion and is frozen in place.

We propose that the double requirement (wh-criterion, V2 requirement on Force) is achieved by an alternative satisfaction of the wh-criterion in FocP. Let us assume that die in Force is not just a neutral head but may also host discourse features, encoded as a D-feature. D-enriched Force spells out as die. Inserting die in Force enriches Force in the same way that adding Φ-features on Fin enriches Fin. In the same way that Φ-enriching Fin allowed for indirect satisfaction of the Subject criterion, we assume that D-enriched Force can satisfy a criterial condition associated with the functional projection it selects. In the example under discussion, D-enriched Force can locally satisfy the wh-criterion in Foc.

In line with Rizzi and Shlonsky, we take it that the D-enriched Force head must be licensed independently; this is achieved by the movement of a discourse-related constituent to SpecForce, in other words, the wh-constituent wanneer moves to SpecForcP where it will satisfy the Force-V2 condition. On its way, the wh-phrase skips FocP; it indirectly contributes to the satisfaction of the wh-criterion because it licenses the D-enriched head Force.

Example (48) reveals an interesting contrast in judgments between the higher acceptability ranking of sentence (48a), which includes a sentence medial deictic adverb dan ‘then’, and the lower ranking of (48b), which lacks this adverb. The impact of the presence of the deictic adverb on the acceptability of the example suggests that the presence of die entails a D-linking effect on the wh-phrase. Assuming that dan ‘then’ anchors the sentence more explicitly to the discourse, the effect noted here leads us to suggest that the presence of die contributes a D-linking effect, in other words that die is specifically compatible with D-linked wh-constituents. This needs further study.

We assume that the D-feature on die is invariably present and encoded in Force, and that the D-enriched Force head in conjunction with the constituent in its specifier will satisfy the criterial condition of the projection it dominates. For speakers rejecting wh-specifiers in die patterns, we would assume that D-enriched Force cannot satisfy the wh-criterion.

With adjunct-initial examples such as (49), two options arise. One is that the initial constituent morgen ‘tomorrow’ could be taken to be topical: it bypasses TopP in (49b), but it indirectly contributes to the satisfaction of the Topic criterion because it licenses the D-enriched head Force. Alternatively, the adjunct morgen does not give rise to a criterial condition and simply moves to SpecForceP to license the D-feature on Force.

Epistemic or speaker-oriented adverbials such as waarschijnlijk ‘probably’ and spytig genoeg ‘sadly’ are compatible with invariant die. These adverbials are not plausible topics, witness the fact that they resist dan-resumption in StD and its varieties, as shown in (28). We assume that in such cases as well, the adverbials merely move to SpecForceP to license the D-feature on Force.

Unfortunately, as it stands, there arises an unfortunate redundancy in our account. SpecForceP has to be filled for two distinct reasons: (i) on the one hand, SpecForceP has to be filled because the Ghent dialect is a Force-V2 language in Wolfe’s typology; (ii) on the other, SpecForceP has to be filled because, when realized as invariant die, Force is D- enriched and its D-feature has to be independently licensed. We need to examine in future work how this redundancy can be eliminated.Footnote ¹⁰

5.2.4. The nature of invariant die

Our reliance on the availability of a D-feature on die raises the question as to the nature of that D-feature. At the moment, our intuition remains rather vague, but we assume the D-feature functions as a discourse connector, explicitly anchoring the sentence into the ongoing discourse. The discourse anchoring function of die might be confirmed by the following anecdotal observation from our informant Professor Emeritus Luc De Grauwe:

‘bij ontmoetingen (met mezelf of in ruimer familieverband) viel soms een keertje een korte stilte in het gesprek/de small talk. Dan had mijn tante de gewoonte, telkens als eerste die stilte te doorbreken door het uitspreken van het enkele woordje dieë (met langgerekte eerste lettergreep) – dit gewoon om het gesprek weer op gang te (laten) brengen, eventueel met een ander onderwerp.’

in the course of meetings (with myself or in the larger family circle), it would happen that a sudden silence occurred in the conversation/small talk. Then my aunt had the habit of being the first to interrupt that silence by pronouncing the word dieë (with long first syllable) – with the sole purpose of getting the conversation going again (possibly on a different topic) (Luc De Grauwe, p.c., 16.08.2017; translation by the authors).Footnote ¹¹

The longer form dieë, which is used in isolation, can be analyzed as the strong form of invariant die, employed in an elliptical context: dieë spells out the head Force of an utterance whose content remains implicit. The use of isolated dieë as a conversational move could be seen as a way of ensuring a minimal illocutionary act. Dieë can be seen as the overt encoding of the speaker’s engagement with the communicative situation and the ongoing discourse, be it as a way of taking or of holding the floor.

Noted exceptions to the V2 generalization are yes/no questions (50a) and imperatives (50b), which display a linear Verb first order. One hypothesis is that such patterns conform to the V2 constraint because a null operator satisfies the Force-V2 condition (50c,d) (cf. Holmberg Reference Holmberg, Kiss and Alexiadou2015, Reference Holmberg, Theresa Biberauer, Wolfe and Woods2020, for relevant discussion).

By our account, constituents qualifying as the first constituent in a V2 pattern appear to the immediate left of invariant die. All things being equal, our analysis might be taken to predict that invariant die is available with yes/no questions and with imperatives, effectively leading to a pattern without an overt antecedent.Footnote ¹² However, the pattern is not attested in the corpus, and yes/no questions and imperatives are judged incompatible with die resumption (51), also in those cases in which what would be a suitable initial constituent is supplied (52):

In an earlier account (De Clercq and Haegeman Reference De Clercq and Haegeman2018), we proposed that resumptive die spells out a declarative Force head – in other words, that it is a declarative root complementizer.Footnote ¹³ This hypothesis leads to at least one empirical problem in that die resumption is compatible with wh-interrogatives. We now propose that the incompatibility between invariant die and imperatives and yes/no questions arises because the relevant null operators are not able to license the D-feature on D-enriched Force.

5.2.5. Force-V2 without die: a first sketch

In the light of our proposals, we also need to re-examine the derivation of regular V2 sentences which do not display invariant die, such as (53a). Derivation (53b) is our first attempt. Assume that wanneer ‘when’ carries the wh-feature. We assume that in die-less varieties of V2, the finite verb moves up to Force. FocP is projected. We take it that V-to Force movement transits through Foc and picks up the Foc-head and the Foc feature encoded on it. The newly created complex head Fin-Foc-Force allows for delaying the satisfaction of the LP criterion at the level of ForceP: (53) exemplifies the derivation.