A Cartographic Approach to Verb Movement and Two Types of FinP V2 in German

Abstract

In this paper, two syntactic configurations are considered that involve V-to-C movement in present-day German: Verb Second in run-of-the-mill declarative clauses and Verb Second in non-assertive embedded contexts. Along the lines of the cartographic approach and on the basis of syntactic and semantic evidence, it is proposed that in both constructs, the finite verb targets neither Force° nor the head of any other projection hosting a moved constituent in its specifier, but, rather, that it moves into the lowest head in the extended CP layer, namely Fin°. As a result of this, (at least) two types of verb raising to Fin° are to be postulated in this language: one that is triggered by discourse/information structure (V2₁) and one that results from mechanical movement to C elicited by an otherwise lacking lexicalization of the relevant left-peripheral head (V2₂).

1. Introduction

Present-day German is a so-called “asymmetric” Verb-Second (henceforth: V2) language. Declarative main clauses exhibit obligatory V-to-C movement, and the topological position to the left of the finite verb (in pre-Rizzian terms: Spec,CP) must be occupied by one XP in order to satisfy an EPP(-like) feature on the left-peripheral projection hosting the verb. In embedded clauses introduced by an overt subordinating conjunction, leftward movement into C is formally blocked by the presence of the complementizer in the same head position standardly targeted by the verb (cf., among many others, Bach 1962; Bierwisch 1963; Den Besten [1977] 1983; Thiersch 1978; Poletto 2002; Adger 2003; Holmberg 2015). These two configurations are illustrated in (1) and (2), respectively:

(1)	a.	Maria	hat	das	Buch	nicht	gelesen.
		Maria	aux.prs.3sg	the-acc.sg	book	neg	ptcp-read-ptcp
		‘Maria has not read the book.’
	b.	[CP Mariax [C° hati [IP/VP tx das Buch nicht gelesen ti]]].
(2)	a.	…dass	Maria	das	Buch nicht		gelesen
		that	Maria	the-acc.sg	book neg		ptcp-read-ptcp
		hat.
		aux.prs.3sg
		‘…that Mary has not read the book.’
	b.	[CP [C° that [IP/VP Maria das Buch nicht gelesen ti [I° hati]]]].

With respect to the operation that triggers the movement of the part of the predicate bearing the tense-aspect-mood features into C° in main clauses, Bayer and Freitag (2020) argue that the lexical component of the finite verb is always interpreted in situ, i.e., in V. In the authors’ approach, verb raising is thus a purely syntactic effect in German and does not result from semantic-feature checking.

In recent years, the idea has consolidated in cartographically oriented syntax that languages with a grammaticalized Verb-Second (V2) rule may fall into the group of so-called “ForceP-V2” or of “FinP-V2” systems (Poletto 2013; Wolfe 2016, 2018; De Clercq and Haegeman 2018).

The former entail a derivation of the type illustrated in (3a), in which the finite verb cyclically moves to Force° after leaving a trace in Fin°—thereby inducing a “bottleneck effect” (Haegeman 1996; Roberts 2004; Cardinaletti 2010; Hsu 2017). The pre-verbal constituent is raised from its middle-field1 base-generation site to Spec,ForceP via Spec,FinP. In the latter position, the moved constituent leaves a trace that blocks any further movement from the IP/VP area, so that this specifier functions as a “bottleneck” filtering the material entering the left edge of the clause. This operationalization is derived from the fact that ForceP-V2 languages are supposed not to allow for any exceptions to the V2 constraint, which is why ForceP, the highest clause-internal projection in the Rizzian (1997) Split CP, is taken to be the landing site of both elements. The term “FinP-V2 language”, on the other hand, refers to a syntactic derivation in which both the verb and the XP in initial position target the lowest left-peripheral projection, FinP, as their landing site (3b). This pertains to the possibility of having V3 or similar patterns involving a larger portion of the C-domain under specific conditions:

(3)	a.	[ForceP XPi [Force° Vfinx] … [FinP ti [Fin° tx [TP ti … tx ]]]]	(ForceP V2)
	b.	[ForceP [Force°] … [FinP XPi [Fin° Vfinx[TP ti … tx ]]]]	(FinP V2)

Differently, e.g., from Old Romance (Salvesen 2013; Poletto 2014; Wolfe 2016, 2018), other Germanic varieties like Cimbrian (Bidese et al. 2012; Cognola 2013) or Kiezdeutsch (Wiese 2012; Freywald et al. 2015; Walkden 2017; Lowell Sluckin and Bunk 2023) or German contact varieties like Iowa German (Rocker 2022), as well as older stages of the German language, which are claimed to exhibit a relaxed (“low”) V2 in most theoretically informed studies of the last decades (cf., e.g., Axel 2007; Petrova 2012; Walkden 2014), contemporary German is generally classified as a strict (“high”) V2 system with a ForceP-V2 rule. This is assumed to be corroborated by the ungrammaticality of linearizations like (4a) and (4b), showing that neither two arguments nor a fronted past participle followed by an argument seem to possibly be hosted in the left periphery:

(4)	a.	*Der	Hans	die	Maria	hat	gestern	getroffen.
		the.nom	Hans	the.acc.sg	Maria	aux.prs.3sg	yesterday	pctp-meet-ptcp
	b.	*Getroffen	der	Hans	hat	die	Maria	gestern
		pctp-meet-ptcp	the.nom.sg	Hans	aux.prs.3sg	the.acc.sg	Maria	yesterday
		(int.:) ‘Hans met Maria yesterday.’2

In this paper, I propose a novel taxonomy of V2 in (present-day) German based on two premises: (i) both distributionally and interpretively, V2 is not a uniform rule, but rather a constraint that is to be defined locally, i.e., there are different types of V2 that have to be related to corresponding phenomena at the syntax-semantics interface; (ii) these different types of V2 follow from different left-peripheral configurations. In particular, I will contend that V2 in German is basically (i.e., with respect to its standard arrangement) a FinP, not a ForceP phenomenon.

In what follows, I will focus on two (of possibly multiple) types of discernible V2 patterns which I contend can be postulated to result from verb movement to Fin°, namely a discourse-related/information-structurally-motivated V2 (V2₁) observable in run-of-the-mill declarative main clauses and a merely syntactic V2 (V2₂) that occurs in some types of asyndetic embedded clauses. The argumentation pursued in this paper is not intended to be exhaustive with respect to the V2 types attested in present-day German, but only refers to the arrangements addressed here.

The paper is structured as follows: in Section 2, it is shown that declarative-main-clause V2 in German cannot be analyzed as a ForceP phenomenon in light of the presence of some surface-Verb-Late(r)3 patterns that suggest finite-verb movement to a much lower head in the CP domain; in Section 3, some arguments in favor of a FinP analysis of one of the V-to-C patterns attested in embedded contexts are discussed. Section 4 concludes and summarizes the results.

2. V2 in Matrix Clauses

2.1. Preliminary Remarks

As observed above, V2 is to be defined as a generalized rule for (declarative) main clauses necessarily involving a combination of leftward movement of the finite verb and cyclic movement of one constituent via Spec,FinP into some left-peripheral specifier in which it appears at PF.4 In this section, it will be proposed that these two operations target, respectively, Fin° and some higher CP specifier depending on the grammatical or information-structural status of the constituent preceding the verb. In other words, main-clause V2 is to be treated as a non-Spec-Head configuration.

Before addressing the relevant data, a theoretical premise needs to be made: in the present account, V2 is not a (primarily) linear constraint, but rather an inviolable structural rule of the grammar of German. The structural and not (only) descriptive nature of the V2 constraint consists of a combination of V-to-C and movement of only one phrase into a dedicated C-specifier, i.e., it entails that multiple XP-raising is excluded but does not impose a ban on generating—under specific conditions—some material spelled out in the sentence directly in the left periphery. Therefore, not all surface word orders of the type “XP > finite verb” in all languages are to be called “V2”. V2 in the strict sense results from the application of a systematic progression of derivational operations invariably leading to one or a (very) limited set of possible left-peripheral word-order options, as is the case in German. The available options must all involve a CP head and the intrasentential domain preceding that position. In this sense, one may assume that two sentences of English like (5) and (6), which are descriptively identical with respect to the lexicalization of their left domains (XP > Vfin), are not even comparable from a structural point of view, i.e., in relation to the operations taking place covertly to derive them. For structures of the type in (5a), the finite verb is standardly assumed to remain in the extended verb projection, with the inflectional features percolating from the higher I° to the VP head (Haegeman 2006b, p. 169) (5b). In fact, the area between the subject and the verb in this sentence could in principle be occupied by further phrases (viz. maximal projections), as exemplarily illustrated in (5c). This is not the case in (6a), where the overt word order unfailingly follows from the application of a rule moving the wh-phrase into some left-peripheral (CP) specifier generally identified with FocP (Rizzi 1997) or whP (e.g., Shlonsky and Soare 2011) and the finite verb being raised to C° ((5b); example from Haegeman and Guéron 1999, p. 172). In the standard case, indeed, no full constituents may surface between a preposed wh-element and the finite verb (6c). Wh-questions are in fact one of the structures of present-day English exhibiting residual V2 (Rizzi 1996):

(5)	Non-V2
	a.	Mary read that book.
	b.	[IP Maryx [I° ] [VP tx never read[+tns/ps/num] that book]]
	c.	Mary definitely never read that book.
(6)	V2
	a.	Whom will Thelma meet after lunch?
	b.	[CP Whomx [C° willi [IP/VP tx meet ti after lunch]]]?
	c.	*Whom definitely will she meet after lunch?

While in languages like English it is the IP layer that is occupied in unmarked declarative clauses of the type illustrated in (5) (i.e., without clause-initial topics, foci, etc.), in German the linearization of matrix structures is primarily a CP phenomenon.

Differently from what the general idea of a restriction bearing this name would suggest (and crucially for the point to be made in the present paper); however, the term “V2” does not imply that the finite verb must necessarily be the second element linearly surfacing in the clause. Rather, it entails that only one XP may (but must) enter—or be merged into—the left periphery of the clause in order to satisfy an EPP(-like) feature requiring that Spec,CP not be empty. In other words, the very implications of V2 are exclusively to be understood in structural, not in linear/descriptive terms. This (not literal) formulation of the constraint can be taken to create at least two tangible expectations:

-

that most main clauses are, with respect to their statistical frequency, de facto linearly V2. Clauses with a V3, V4, etc., word order, if possible in a given configuration, will represent exceptions within their corresponding rule: if there is no specific interface condition licensing the presence of additional material in the C-domain, the clause will have an overt V2 arrangement;

-

that Verb-Late(r) patterns may still be possible under the condition that they not violate the constraint imposed by the combination of EPP and bottleneck effect.

In the next paragraphs, it will be shown that these expectations are, in fact, mirrored by the data and that this is a key argument in favor of a formalization of German (matrix-clause) V2 as a FinP phenomenon.

2.2. What Verb-Late(r) Tells Us about Verb Movement

Declarative-clause word orders in which the finite verb is preceded by more than one element are possible and productive in German (among many others, Grewendorf 2002a, 2002b; Müller 2003; 2005, p. 153; Pasch et al. 2003; Frey 2004, 2005; Breindl 2008, 2011; Speyer 2008; Volodina and Weiß 2010; Catasso 2015; 2021, p. 154; Métrich and Courdier 1995; Speyer and Weiß 2018; Samo 2019; Breitbarth 2018, 2022, 2023; Bunk 2020; Wiese et al. 2020, p. 155; Lowell Sluckin and Bunk 2023).

One of the main problems concerning the theoretical treatment of such patterns in the literature is that the different phenomena are mainly addressed individually and/or based on very different premises. This is due to the fact that the development of an all-in-one solution to account for the resulting arrangements necessarily implies a number of different postulations about the nature of the left periphery of German (e.g., concerning the interaction between Merge and Move operations) that need to be accepted at once.

In (7), some of these structures apparently violating the V2 constraint addressed above are exemplified:

(7)	a.	Den	Hans,	den	mag	jeder.
		the-acc	Hans	res.acc.sg	like-prs.3sg	everyone-nom
		‘Everybody likes Hans.’
		(Frey 2004, p. 204)
	b.	Wenn	es	regnet,	dann	bleiben	wir	zu	Hause.
		if	expl	rain-prs.3sg	then	remain-prs.1pl	we-nom	at	home
		‘If it rains, we will stay at home.’
		(Reis and Wöllstein 2010, p. 115)
	c.	Context:
		In Chile gebe es, so fuhr er fort, ein System, “das diktatorisch, aber per definitionem vergänglich ist”.
		‘In Chile, there is—he continued—a system that is dictatorial but, by definition, ephemeral.’
		Relevant sentence:
		In	seiner	Heimat	hingegen	gebe	es	eine
		in	his-dat.sg	homeland	instead	give-sbj.prs.3sg	it-nom.sg	a-acc.sg
		Diktatur	ohne	jedes	Element	der	Hoffnung.
		dictatorship	without	every-acc.sg	element	the-gen.sg	hope
		‘In his homeland, instead, there is a hopeless dictatorship.’5
		(Breindl 2011, p. 42)
	d.	Wenn	was	passiert,	man	geht	nach	Connewitz.
		if	something-nom	happen-prs.3sg	one-nom	go-prs.3sg	to	Connewitz
		‘If something (bad) happens, you’ll automatically think of Connewitz.’
		(Breitbarth 2022, p. 15)
	e.	Der	Junge,	sobald	er	den	Alten	nur
		the-nom.sg	boy	when	he-nom	the-acc.sg	old-acc.sg	prt
		verstanden	hatte,	nickte	und	sprach:	O ja!	Sehr	gern.
		understand-ptcp	aux.pst.3sg	nod-pst.3sg	and	say.pst.3sg	oh yes	very	gladly
		‘When he understood the old man, the boy nodded and said: “Oh yes, I’d love to.”’
		(Lee 1975, p. 120)
	f.	Am	Sonnabendmorgen	im	frühesten	Zug	war	er
		on-the-dat.sg	Saturday-morning	in-the-dat.sg	earliest-dat.sg	train	be-pst.3sg	he-nom.sg
		unausgeschlafen	genug,	das	ganze	Unternehmen	zu	verfluchen.
		sleep-deprived	enough	the-acc.sg	whole-acc.sg	company	to	curse-inf
		‘On Saturday morning, on the earliest train, he was sleep-deprived enough to curse at the whole company.’
		(Van de Velde 1978, p. 134)

In (7a), illustrating a left dislocation, a fronted DP is taken up by a d-pronominal resumptive bearing the same φ-features as the dislocate occurring preverbally, i.e., in the so-called “prefield”.6 Although this is clearly a correlative construction, the two left-peripheral elements appearing to the left of the finite verb do not form a syntactic unit at PF. The structure in (7b), in which an utterance-initial conditional clause is resumed by preverbal dann (‘then’), can be characterized—mutatis mutandis—as an adverbial counterpart to (7a). Of course, the type of “agreement” observable between the topicalized XP, which functions as a clausal frame setter here insofar as it circumscribes the frame (one that is projected into a non-actual world) in which the content of the clause is to be interpreted, and the resumptive does not involve any case and number, but these two elements are referentially identical.7 In (7c), a contrastive PP frame setter is linearly followed by a so-called “post-initial marker” (Métrich and Courdier 1995; Pasch et al. 2003; Breindl 2008; Volodina and Weiß 2010; Karagjosova 2012; Lenker 2014; Catasso 2015; Speyer and Weiß 2018) in the prefield. Post-initial markers are originally adverbial, particle-like elements that systematically surface in second position and accompany a fronted full XP with the function of lexicalizing their information-structural status. In the relevant sentence, the contrastive topic in seiner Heimat ‘in his homeland’ is opposed to the contrastive topic in Chile ‘in Chile’, which is introduced in the previous sentence, and the marker hingegen ‘instead’ has a contrastive value, its presence being exclusively licensed by a contrastive XP to its left. In (7d), a frame-setting topic precedes a pronominal subject in the left periphery. This configuration has so far been described in detail in a large amount of literature on urban varieties of German and other V2 languages (for German/Kiezdeutsch, cf., among many others, Wiese 2012; Walkden 2017; Wiese et al. 2020; Lowell Sluckin and Bunk 2023). The pattern, however, while not being very frequent (cf., e.g., the estimates by Speyer (2008) and Wiese and Rehbein (2016), which point to a rate below 0.05%, at least in the consulted corpora), is still productive in spontaneous spoken interaction. Especially in the very last years, a number of researchers have discussed V3 orders of this type in spoken German that do not seem to result from the influence of a multilingual context or language contact (Bunk 2020; Catasso 2021; Breitbarth 2022, 2023). Note that this wenn-clause—which can be interpreted conditionally or temporally here—is interpreted along the lines of (7b) in that it introduces a condition to or temporal frame for the validity of the content of the main clause. In (7e), a clausal frame setter appears between a topic which is also the subject of the utterance and the finite verb. In (7f), two non-clausal frames appear in the prefield. Note that even though am Sonnabendmorgen ‘on Saturday morning’ and im frühesten Zug ‘on the earliest train’ can be assumed to concur to form a “big frame” within which the state/action described in the sentence is to be interpreted, they belong to different semantic clines: the former realizes a temporal, the latter a local reference.8

These patterns already suggest that even for V2 languages like German, a Split-CP analysis is possibly more adequate than one in which the CP consists of one projection whose specifier can only be occupied by a single XP. In principle, one possibility to make sense of these data would be to analyze constructions like those in (7a), (7b), and (7d) as involving extrasentential Merge of the XP in first position (i.e., base-generation of the highest full constituent in a domain that does not interact with the syntactic computation of the clause-internal area). The structure in (7f) could be treated as an instantiation of a complex PP assembled by adjunction of the right XP (here: im frühesten Zug) to the left constituent (here: am Sonnabendmorgen). However, one would have to admit that at least the configurations in (7c) and (7e) cannot be simply dismissed as “exceptions”, since they exemplify quite productive patterns both of colloquial spoken and written Standard German and the first element in their linearization is an argument of the lexical verb which cannot plausibly be assumed to be first-merged clause-externally.

More interesting than the single patterns, however, is the occurrence of multiple phenomena in the same sentence, which has been mainly neglected in the literature on present-day German so far (for some exceptions, see Speyer and Weiß 2018; Catasso 2021; Breitbarth 2022). This interplay provides additional information not only with respect to a multiple-specifier analysis of the preverbal area, but also—crucially for the scope of the present work—in relation to the actual position of the finite verb in the left periphery of the main clause.

In his model of the CP domain (8), Rizzi (1997) assumes that the left (pre-IP) edge of the sentence consists of an array of highly specialized projections that host or lexicalize grammatical features such as interrogativity or information-structural categories such as Focus and Topic (the latter being syntactically recursive; also cf. Frascarelli and Hinterhölzl 2007; Hinterhölzl 2017). The CP area is introduced by ForceP, which encodes clause type and illocution, and is delimited in the immediate pre-middle-field area by FinP, which is responsible for finiteness:

(8)	ForceP > TopP* > FocP > TopP* > FinP > IP …
	(adapted from Rizzi 1997, p. 297)

2.3. Main-Clause V2 as V-to-Fin

In the approach advocated in this paper, an optimal way to account for the data in (7) is assuming that the constituents surfacing in these patterns are positioned in the clause-internal portion of the left periphery (i.e., below ForceP) at PF and the finite verb is pronounced no higher than Fin° in a model combining the presence of EPP/bottleneck and a Split CP that can be summarized as follows:

▪

Left dislocations sensu lato (that is, constructions like (7a) and (7b)) are derived by cyclic leftward movement of the topic (disl in the representation below) from its base-generation site in the middle field of the clause into the specifier in which they surface (one that is compatible with the interpretive features of the constituent, e.g., with contrastive or aboutness topicality) (cf. Cinque 1977 and Grewendorf 2002a). This cyclic raising operation involves an intermediate pit-stop in Spec,FinP, where the XP leaves a trace that blocks any further movement into the left periphery. In the cases illustrated in (7), which exhibit linear V3, the trace in Spec,FinP is pronounced (i.e., it is externalized as a trace spell-out that has the phonetic substance of a resumptive) (9a). The spell-out of the resumptive (res in the representation below) element, however, is optional: if this element is silent, there are no consequences whatsoever for the grammatical acceptability or the semantic interpretation of the sentence. The resumptiveless (that is, non-redundant) version of the sentence is the one commonly occurring in the standard language (9b)–(9c). This analysis is in line with the idea sketched above that the principle underlying the bottleneck effect is inviolable and that this type of linear V3 in fact corresponds to a structural instantiation of V2:9

(9)	a.	[ForceP [TopP [disl]x … [FinP [res]x [Fin° Vfini [IP tx … ti ]]]]]
	b.	Den	Hans	mag	jeder.
		the-acc	Hans	like-prs.3sg	everyone-nom
		‘Everybody like Hans.’
	c.	Wenn	es	regnet,	bleiben	wir	zu	Hause.
		if	expl	rain.prs.3sg	remain-prs.1pl	we-nom	at	home
		‘If it rains, we will stay at home.’10

▪

(Both DP and adverbial) left dislocation can co-occur with the pattern in (7c). Although some analyses propose that contrastive markers like hingegen in (7c) are part of the XP topicalized into the left periphery (cf., e.g., Volodina and Weiß 2010), more recent studies have bolstered the hypothesis that such elements are base-generated in the head position of some left-peripheral projection hosting a corresponding topic (Speyer and Weiß 2018; Catasso 2015, 2021) (“KP” in the example below).11 If this idea is on the right track, neither post-initial markers per se nor a combination of left dislocation and topic particles like hingegen violate the V2 constraint, since only one XP is moved into the left periphery. Also, in this case, the verb cannot move past Fin°. This configuration is shown in (10):

(10)	a.	Context
		Ich	finde	alle	Kollegen	sehr	sympathisch.
		I-nom	find-prs.1sg	all-acc.pl	colleague-acc.pl	very	friendly
		‘I find all my colleagues very friendly.’
		Relevant clause
		Den	Hans	aber/hingegen/jedoch,	den	mag	ich	nicht.
		the-acc.sg	Hans	however/instead	res.acc.sg	like-prs.1sg	I-nom	neg
		‘Hans, instead—I don’t like him.’
	b.	[ForceP [KP [disl]x [K° marker] [FinP [res]x [Fin° Vfini [IP tx … ti ]]]]]
		base generation

▪

The data in (7d) and (7e), in which an adjunct and an argument co-occur in the preverbal domain, suggest that (i) the extended CP of German includes at least two positions in which frame setters may surface, and (ii) a frame setter can be moved (like in (7b)), but it can also be base-generated in some kind of Last Resort configuration that prevents the structure from crashing by violating the V2 constraint, as in these two examples. In (7d)–(7e), this is apparent because the other XP in the prefield is necessarily the one interactive with the V2 syntax of the clause (in both cases, it is the subject of the clause) and because neither wenn- nor the sobald-clause can leave a trace in Spec,FinP.12 What is more, the adjuncts are separated from the rest of the material in the CP domain by a comma intonation (which is also empirically confirmed in Breitbarth’s (2022) analysis for structures like (7e)). This position is, in fact, already occupied by the other constituent (as is arguably the case in (7d)) or by its trace (as in (7e)), and as outlined above, only one XP can enter the left periphery. The topological positions in which the adjuncts in these two examples surface are situated above and below (one of) the positions hosting topics, respectively. The higher of these positions, indeed, is linearly followed by an indefinite pronoun (i.e., by a non-topical, phonologically weak element that plausibly surfaces in Spec,FinP) in (7d), but can also precede full-XP topics, which are arguably hosted in one of the Spec,TopPs also independently proposed by Rizzi (1997).13 In (11), these positions are neutrally labeled “YP” and “ZP”. By assuming this derivation, the force of the bottleneck restriction is preserved, since only one XP has entered the left periphery in the configuration assumed here. Again, the most immediate consequence with respect to the position of the finite verb is that this element must surface in Fin° in the overt syntax. Also note that in both structures, the trace spell-out of the argumental topic can optionally be pronounced in Spec,FinP (cf. (12a)–(12b), in which two of the corpus examples mentioned so far have been accordingly modeled):14

(11)	[ForceP [YP Adjunct [TopP Topicx [ZP Adjunct … [FinP tx/resx [Fin° Vfini ] [IP tx … ti ]]]]]]                      base generation    base generation

(12)	a.	Heute,	die	Goethe-Institute,	die	heißen	alle	Max Müller Bhavan
		today	the-nom.pl	Goethe-institute-nom.pl	res.nom.pl	be-called-prs.3pl	all-nom.pl	Max Müller Bhavan
		‘Today, all Goethe institutes are called Max Müller Bhavan’
	b.	Der	Junge,	sobald	er	das	verstand,	der	nickte …
		the-nom.sg	boy	when	he-nom	that-acc	understand-pst.3sg	res.nom.sg	nod-pst.3sg
		‘When the boy understood that, he nodded …’

▪

As to the pattern presented in (7f), where two distinct but semantically compatible adjuncts linearly precede the finite verb, it can be assumed that these are assembled into a complex XP prior to movement to the left periphery, but that they preserve their plasticity within this phrase in that they can in principle be separated—if necessary—once they have reached the CP domain. In fact, the left-peripheral “big frame” in a sentence like (13a) can, for instance, and given the appropriate context: (i) be “interrupted” by an intervening post-initial particle (13b); (ii) precede as a whole a post-initial particle (13c); (iii) be taken up by a preverbal adverbial resumptive (13d); and (vi) be resumed as in (13d) with an intervening (13e) post-initial marker or a post-initial particle following the XP (13f):

(13)	a.	Gestern	im	Bus	saß	ich	Colin Farrells	Bruder	gegenüber.
		yesterday	in-the-dat.sg	bus	sit-pst.1sg	I-nom	Colin Farrell-gen	brother	v.prt
	b.	Gestern	aber	im	Bus	saß	ich	C.F.s	Bruder	gegenüber.
		yesterday	however	in-the-dat.sg	bus	sit-pst.1sg	I-nom	C.F.-gen	brother	v.prt
	c.	Gestern	im	Bus	aber	saß	ich	C.F.s	Bruder	gegenüber.
		yesterday	in-the-dat.sg	bus	however	sit-pst.1sg	I-nom	C.F.-gen	brother	v.prt
	d.	Gestern	im	Bus,	da	saß	ich	C.F.s	Bruder	gegenüber.
		yesterday	in-the-dat.sg	bus	res	sit-pst.1sg	I-nom	C.F.-gen	brother	v.prt
	e.	Gestern	aber	im	Bus,	da	saß	ich	C.F.s	Bruder	gegenüber.
		yesterday	however	in-the-dat.sg	bus	res	sit-pst.1sg	I-nom	C.F.-gen	brother	v.prt
	f.	Gestern	im	Bus	aber,	da	saß	ich	C.F.s	Bruder	gegenüber.
		yesterday	in-the-dat.sg	bus	however	res	sit-pst.1sg	I-nom	C.F.-gen	brother	v.prt
		‘Yesterday on the bus, (however,) I sat in front of Colin Farrell’s brother.’
		(sentence in (13d) adapted from: sarahwundermann.ch (2010) (blog))

If the assumptions about the makeup of the left periphery of German made so far are correct, these data suggest that in contexts like (13), the “big frame” gestern im Bus (‘yesterday in the bus’) is moved from the middle field into the left periphery, thereby leaving a trace in Spec,FinP that can be either pronounced ((13d), (13e) and (13f)) or silent ((13a) and (13c)), and reaches its surface position either as a block (arguably in (13a), (13c), (13d) and (13f)) or in a “split” configuration ((13b) and (13e)) in which, I suggest, only the component of the “big frame” that is to be interpreted contrastively is further moved to the higher specifier in whose head aber (‘however’) is generated, while the lower, presumably less salient part im Bus is “stranded” in an intermediate Top-position between the higher specifier and FinP. The higher and the lower specifier can be preliminarily identified as the YP and ZP illustrated in (11) above. Thus, the projections labeled “YP” and “ZP” here are in principle both able to host base-generated or moved constituents given the relevant context:15

(14)	[ForceP [YP [gestern]y [Y° marker] … [ZP [ty im Bus]x [FinP tx/resx [Fin° Vfini] [IP tx … ti ]]]]]                      base generation

I take the data presented so far to provide strong evidence in favor of the assumption that the finite verb does not move to Force° in run-of-the-mill (declarative) main clauses, but occupies a lower, pre-IP head position. This is due to the fact that the single V3 patterns addressed in this section and in particular the simultaneous co-occurrence of multiple Verb-Late(r) phenomena in one and the same sentence necessitate more than one intrasentential preverbal specifier for the structure to be derived correctly. The very position in which the finite verb surfaces, I suggest, cannot be one of the heads between FinP and ForceP (i.e., e.g., Top° or Foc°), either. This is quite apparent if one considers that the only sensible configuration that would allow this assumption is one in which the finite verb is attracted to the head of the projection hosting a moved constituent in its specifier, since this XP is supposed to be the one interacting with the V2 syntax of the clause. However, the relative variation observable in the CP domain of German makes this hypothesis very implausible. According to this idea, one would expect the verb to be raised into the head of the TopP hosting der Hans in a sentence like (15). The actual linearization of the utterance, however, is not compatible with this derivation, since der Hans occurs in a sequence in which it is followed by other elements correlatively bound to it (here, the contrastive marker aber and the resumptive der) or independently occurring in the left periphery (here, the sobald-adjunct):

(15)	Der	Hans	aber,	sobald	er	das	hörte,
	the-nom.sg	Hans	however	when	he-nom	that.acc	hear-pst.3sg
	der	sagte
	res.nom.sg	say-pst.3sg
	‘However, Hans, when he heard that, said …’

Assuming that the finite verb occupies Force° in a structure like (15) amounts to implying that the remaining left-peripheral elements are either all positioned in Spec,ForceP—a possibility that can easily be ruled out a priori—or that some or all of them are situated clause-externally (above ForceP). The latter option is not very appealing considering that der Hans, the XP surfacing in the highest position, is the subject of the clause and must therefore have been moved into the specifier in which it appears.

I conclude from this that in (declarative) main clauses, the finite verb systematically occupies Fin° and cannot be moved past that position.16 All patterns illustrated in this section are—structurally—V2 word orders that are realized in different linear variants: V3, V4, and so on. V2 as a structural constraint is derived by (cyclically) moving the finite verb from V° to Fin° and (cyclically) raising one XP into one of the specifiers in the C-domain via Spec,FinP, thereby satisfying the EPP(-like) feature of Comp and correctly deriving the bottleneck effect of V2 languages like German. All Verb-Late(r) patterns can be analyzed as implying a combination of Merge and movement that produces the observable configurations.

After all, it comes as little surprise that Fin° is assumed to host the verb: this is the projection primarily dedicated to the encoding of finiteness in Rizzi’s model, one that contains “a tense specification which matches the one expressed on the lower inflectional system” (Rizzi 1997, p. 283; also cf., among others, Haegeman 1996; Roberts 2004). This is corroborated by the observation that multiple elements can surface in the preverbal domain of a main clause, as illustrated above. The postulation that the verb is moved into Force° in declarative clauses would lead to the (undesirable) conclusion that two linearizations like those in (13a) and (13f) should correspond to radically different underlying structures although they are grosso modo semantically equivalent if one abstracts away from the information-structural details that may minimally differentiate them from each other.

Moreover, ForceP can be activated independently of its lexicalization. If we assume that Force hosts an operator in main clauses defining the clause-typing and illocutionary attributes of a sentence, this projection need not be the target of explicit V-movement. In this sense, the Verb-Late(r) patterns investigated in this section provide a third factor corroborating this assumption. Note that this does by no means imply that ForceP V2 does not exist in any language system, but only that present-day German declarative clauses do not represent an ideal context for this kind of derivation. In (16), a simplified model of the left periphery of German (adapted to Rizzi’s (1997) seminal proposal) is sketched according to the assumptions made here. In ForceP, clause type and illocution are encoded; in the specifiers of the projections labeld “YP”/”KP” in (16), contrastive topics and adjuncts are hosted when the sentence is pronounced; in the corresponding head position, a post-initial marker can optionally be spelled out; Spec,TopP is the landing site of non-contrastive topics, Spec,WhP of interrogative constituents; the specifier of the projection labeled “ZP” hosts non-contrastive adjuncts; SpecFinP functions as the bottleneck of the left edge of the clause and as the locus of the EPP; its head is the position into which the finite verb is moved:

(16)	[ForceP [YP/KP [Y°/K°] [TopP [WhP … [ZP [FinP [Fin°] IP/VP … ]]]]]]

In sum, in this section it has been proposed that V2 is a FinP phenomenon in declarative (and other types of) main clauses. In what follows, a second type of (FinP) V2 will be taken into consideration to show that also in (some) embedded contexts, the idea of V-to-Fin movement seems to be more adequate than what V-to-Force raising would imply.

3. V2 in Embedded Argument Clauses

3.1. Selecting Predicates and Subordinate V2

In German, subordinate structures that realize an argument of the main-clause lexical verb are typically introduced by the complementizer dass (‘that’) (17a). At least some of these structures can be assumed to be both syntactically and semantically dependent on their matrix (cf., e.g., Hooper and Thompson 1973; Haegeman 2006a), i.e., to lack any assertive potential. The presence of a subordinating conjunction in C° blocks the movement of the finite verb into the same position (17b). The complementary distribution of Comp and Vfin is, in fact, one of the main tenets of the topological description of the syntax of German (cf., among many others, Den Besten [1977] 1983). In some cases, however, complementizer deletion is allowed, which in turn licenses (in fact, forces) V-to-C movement (17c). This provides strong evidence in favor of the assumption that the complementizer occupies—if spelled out—the same head position that is targeted by verb movement:

(17)	a.	Hans	meinte,	dass	das	verboten	ist    /	sei.
		Hans	say-pst.3sg	that	that-nom	forbid-ptcp	be-prs.3sg	be-sbj.prs.3sg
	b.	*Hans	meinte,	dass	das	ist   /	sei	verboten.
		Hans	say-pst.3sg	that	that-nom	be-prs.3sg	be-sbj.prs.3sg	forbid-ptcp
	c.	Hans	meinte,	das	ist /	sei	verboten.
		Hans	say-pst.3sg	that-nom	be-prs.3sg	be-sbj.prs.3sg	forbid-ptcp
		‘Hans said that it was forbidden.’

In the (grammatical) examples in (17), the finite verb can be in the indicative or (here: present) subjunctive both in clause-final and in clause-initial position. The latter can be taken to be (a remnant of) a correlate of subordination in these contexts, i.e., the licensing of the subjunctive in these subordinate clauses, which realize the direct object of the verb meinen (‘say’, ‘believe’), indicates that the structure is deeply embedded into the selecting predicate. In the same environments, the indicative is also possible as an alternative to replace the subjunctive, especially in spoken usage, which is not very productive in conceptually oral texts. However, where the subjunctive is at least possible, the content of the corresponding subordinate clause exhibits no illocutionary force of its own.17 For instance, in the context presented above, the embedded predicate <be forbidden> can by no means be interpreted as an assertion, a direct question or an imperative produced by the speaker: the only possible reading is one in which the content of Hans’ (a third person’s) saying/thinking is reported.

While it is true that only the absence of an overt Comp can license V-to-C movement, the alternation between verb-final (henceforth: VF) and V2 word order is not systematic. In fact, the conditions regulating the possibility of having V2 are somewhat fuzzy in present-day German, and the acceptability judgments are in part subject to inter-speaker variation and diachronic change (for an overview of the relevant contexts, which will not be reviewed in detail here, cf., e.g., Dal 1966; Dunbar 1985; Reis 1997; Auer 1998; Freywald 2016).

In general, the V2 pattern illustrated in (17c) tends to be possible if the argument clause complements a verbum dicendi/sentiendi/cogitandi or a doxastic predicate realizing a representative speech act, especially when the content of a conversation is reported. In this case, the complement clause is not (or must not be) asserted,18 and the subject of the matrix predicate can coincide with the speaker (18) or be another individual (19). Indeed, the same pattern is generally allowed if the argument clause complements one of the above-mentioned predicates occurring in a non-assertive (e.g., interrogative, imperative, negative or conditional) main clause (20). The same pattern is also generally possible if the V2 clause is the complement of a DP derived from one of these verbs (21). Moreover, V-to-C movement tends to be licit if the selecting predicate is a (non-modal) desiderative verb or noun (22) or any nominal expression not morphologically derived from any of the verbs mentioned so far, whose reference is projected into another world (23). Both variants of the complement clause given in (24a)–(24b), which are presented in this compact format for space reasons (the relevant punctuation option at the end of each of the sentences is to be referred to the corresponding main-clause governor), could complement any of the matrix predicates in (18)–(23) and exhibit a finite present subjunctive here for illustrative purposes. As noticed above, the corresponding forms can be substituted with a present indicative. A V2 counterpart to the “dass + VF word order” variant is ruled out when the main clause contains a factive predicate (25):

(18)	Da meinte ich zu ihm,/ich bin der Ansicht,/ich habe gehört,/ich vermute, …
	‘Then I said to him/I am of the view/I heard/I assume …’
(19)	Sie sagte,/Maria war davon überzeugt,/er dachte,/er glaubte, …
	‘She said/Maria was convinced/he thought/he believed …’
(20)	Meinst du,/finden Sie nicht,/wenn ich gedacht hätte, …
	‘Do you believe/don’t you find/if I had thought …’
(21)	der Glaube,/der Gedanke/die Annahme/die Vermutung, …
	‘the belief/the thought/the assumption/the conjecture …’
(22)	Ich hoffe,/die Hoffnung,/ich wünschte, …
	‘I hope/the hope/I wish …’
(23)	die Lüge,/die Idee, …
	‘the lie/the idea …’
(24)	a.	VF-variant
		…	dass	das	ein	Fehler	gewesen	sei
			that	that-nom	a-nom	mistake	ptcp-be-ptcp	be-sbj.prs.3sg
	b.	V2-variant
		…	das	sei	ein	Fehler	gewesen
			that-nom	be-sbj.prs.3sg	a-nom	mistake	ptcp-be-ptcp
		‘…that it was a mistake.’
(25)	*Ich	bereue,/	*Hans	bereute,	ich/er	habe	Maria	geholfen.
	I-nom	regret-prs.1sg	Hans	regret-pst.3sg	I/he-nom	aux.sbj.1/3sg	Maria	ptcp-help-ptcp
	(int.:)	‘I regret/Hans regretted helping Mary out.’

At this point, the question arises as to how the V2 pattern presented in this section can be syntactified and how the derivation is to be integrated into the Split-CP model of the left periphery of German addressed above. Section 3.2 outlines the definitory features of the category of object clauses examined here, and Section 3.3 proposes a structural analysis along these lines.

3.2. Zooming In: Focus on Non-Assertive Object-Clause V2

In order to define the limits of the present investigation, it must be pointed out, as already mentioned above, that not all argument clauses that allow a “dass + verb final”/“∅-V2” alternation in German are non-assertive. In the examples given in Section 3.1 ((17)-(24)), the embedded clause has no illocutive potential whatsoever, which licenses the subjunctive. In some cases, however, it is possible to assume that an argument clause may realize a speech act (Kayne 2008; Arsenijević 2009; Müller 2021, 2023). This depends on the context (see, e.g., Fn. 17) and/or on the nature of the selecting predicate. If it is the case that a matrix clause (DP) subordinating a dass- or V2-clause is compatible with the embedding of an assertive reading, the resulting structure will exhibit two fundamental correlates: (i) the verb in the argument clause cannot be in the subjunctive (26); (ii) V2 is virtually always possible, and the clause allows—given the relevant context—for main-clause phenomena such as left dislocation (27a), the licensing of contrastive markers (27b) and combinations of multiple phenomena (27c):

(26)	Man muss wirklich sagen,/ich kann nur bestätigen,/das Problem/die Sache ist, …
	‘I just have to say/I can only confirm/the problem/the thing is …’
	a.	…	dass	die	Hotline	der	Firma
			that	the-nom.sg	hotline	the-gen.sg	company
			furchtbar	ist/*sei.
			terrible	be-prs.3sg/be-sbj.prs.3sg
			(VL)
	b.	…	die	Hotline	der	Firma	ist/*sei
			the-nom.sg	hotline	the-gen.sg	company	be-prs.3sg/be-sbj.prs.3sg
			furchtbar.
			terrible
			(V2)
			‘…the company’s hotline is terrible.’
(27)	Man muss wirklich sagen,/ich kann nur bestätigen,/die Sache ist, …
	‘I just have to say/I can only confirm/the thing is …’
	a.	…	diese	Hotline	hingegen	ist	einfach	super.
			this-nom.sg	hotline	instead	be-prs.3sg	just	great
	b.	…	diese	Hotline,	die	ist	einfach	super.
			this-nom.sg	hotline	res.nom.sg	be-prs.3sg	just	great
	c.	…	diese	Hotline	hingegen,	die	ist	einfach	super.
			this-nom.sg	hotline	instead	res.nom.sg	be-prs.3sg	just	great
			‘…that this hotline (, instead,) is just great.’

Crucially, these phenomena lead to ungrammaticality in syntactic contexts like those addressed in the previous section, where the selecting predicate is not compatible with an embedded assertion. The grammaticality of these sentences would not improve if the verb form exhibited indicative morphology. The relevant data are shown in (28):

(28)	Ich hoffe,/die Idee,/ich würde nicht sagen,/wenn ich gedacht hätte,/Hans dachte, …
	‘I hope/the idea/I would not say/if I had thought/Hans said/believed …’
	a.	…	diese	Hotline	{*hingegen}	sei	besser.
			this-nom.sg	hotline	instead	be-sbj.prs.3sg	better
	b.	…	diese	Hotline	{*, die}	sei	besser.
			this-nom.sg	hotline	res.nom.sg	be-sbj.prs.3sg	better
	c.	…	diese	Hotline	{*hingegen, die}	sei	besser.
			this-nom.sg	hotline	instead   res.nom.sg	be-sbj.prs.3sg	better
			(int.:) ‘…that this hotline (, instead,) is better.’

In consideration of these facts, it is reasonable to conclude that: (i) the derivation of the linear arrangements in the V2 examples in (26) and (27) and that of the examples discussed in Section 3.1 must be different. In the present paper, the former, which seems to make use of a larger portion of the left periphery, will not be further considered. The task of formalizing the corresponding derivation is left to future research, and (ii) V2 is not to be conceptualized as a main-clause phenomenon tout court, but rather needs to be derived locally (viz. in divergent ways in different configurations). It is also possible that these two types of embedded V2 are not the only that exist in the system.

In Section 3.3, it will be proposed that the non-assertive V2 pattern(s) described in Section 3.1 provide a further instantiation of FinP V2, but that the linearization of this structure follows from different premises from those debated in the previous parts of the paper.

3.3. Non-Assertive Embedded-Clause V2 as V-to-Fin

In Section 2, it has been contended that the various Verb-Late(r) constructions that are possible in declarative main clauses are not compatible with the idea of a one-specifier-one-head left periphery, which precludes a derivation of V2 with the verb targeting Force° (irrespective of how the respective analysis is technically implemented). In those structures, I suggested, ForceP is activated, but not filled with lexical material.

In non-assertive embedded-clause V2, instead, the position immediately preceding the finite verb can de facto host one and only one element. This also holds for non-phrasal items that have been analyzed as heads here (e.g., post-initial particles). Thus, at least in principle, an analysis in which the verb is cyclically moved into Force° and the only remaining intrasentential position to the left of this head, namely Spec,ForceP, is occupied by the constituent surfacing clause-initially is not excluded by purely syntactic (i.e., linearization-based) evidence. However, I will refute this idea here.

The main argument against a ForceP-V2 configuration for these data comes from the fact that these non-asserted embedded clauses clearly do not display any kind of (even weak or pseudo-) illocutionary independence or potential. The V-to-C movement observable in these patterns only depends on the fact that the system does not allow for finite syntactic structures that—simultaneously—are not introduced by a complementizer and exhibit a linear VF arrangement. This amounts to the well-known generalization by Den Besten ([1977] 1983) that Comp and the finite verb basically compete for the same position.

I assume the structures exemplified in (29) to be derived as in (30). I propose that the complementizer dass is merged in Fin° (and not in Force°) in German, (29a)–(30a) which is also in line with the fact that in a number of West-Germanic (including some German) varieties, subordinating conjunctions forcing a VF word order (may) display overt finiteness morphology matching (or replicating) the features on the verb in final position.19 If the complementizer is not spelled out (as in (29b)), the head of FinP (obligatorily) attracts the finite verb, which in turn activates the raising of an XP to occupy the pre-finite position (namely Spec,FinP). Thus, the finite verb and the constituent in first position occur in a Spec-Head configuration (30b). This operation is information-structurally vacuous, i.e., its only function is the satisfaction of a formal-syntactic requirement (in this case, the EPP-like feature on C) whose violation would cause the derivation to crash. This assumption entails that both in the dass-VL (30a) and in the V2-variant (30b), the main-clause predicate (in the following examples: m.c.) selects a complement clause (i.e., in neutral terms, a CP) that is truncated at FinP, ergo lacks all projections above the one in which the complementizer is base-generated (in dass-clauses) or, in the absence of a complementizer, into which the finite verb is moved:

(29)	a.	VF-variant
		Mir	wurde	gesagt,	dass	das	kein
		I-dat	aux.pst.3sg	ptcp-say-ptcp	that	that-nom	no-nom.sg
		Problem	sei/ist.
		problem	be-(sbj.)prs.3sg
	b.	V2-variant
		Mir	wurde	gesagt,	das	sei/ist	kein	Problem.
		I-dat	aux.pst.3sg	ptcp-say-ptcp	that-nom	be-(sbj.)prs.3sg	no-nom.sg	problem
		‘They said that was not a problem.’
(30)	a.	VF-variant
		[CP [m.c.] [ForceP [TopP [FocP/whP [FinP [Fin° dass [IP XP…Vfin ]]]]]]]
	b.	V2-variant
		[CP [m.c.] [ForceP [TopP [FocP/whP [FinP [XP]i [Fin° Vfiny [IP ti…ty ]]]]]]]

This type of V2 can be assumed to result from an (at least partial) instantiation of so-called “meaningless movement” (Cinque 2018), i.e., of movement yielding effects on PF, but no clear effects on LF: the verb is mechanically attracted to Fin° by virtue of a missing lexicalization of this head, and the EPP determines a further obligatory movement operation of a full phrase from the IP/VP area for formal reasons. Note that applying the idea that some operations are not directly or entirely motivated by feature checking is not an ad-hoc solution here: in fact, if Fin° generates a Comp, the finite verb and the XP would obligatorily surface in some lower position.20

4. Conclusions

In this paper, I have considered two types of leftward-verb-movement configurations in present-day German (declarative-clause V2 and non-assertive embedded-clause V2) and proposed that both phenomena, while being triggered by different mechanisms, result from V-to-Fin raising. The proposed taxonomy is based on two tenets: (i) that V2 as a linear pattern is not the (by)product of a uniform structural constraint, but one to be defined locally, viz. with respect to a single configuration or a group of configurations at some interface (syntax-semantics, syntax-pragmatics, etc.); and (ii) that these putative different types of structural V2 (intended as a sequence of operations in Narrow Syntax that lead to a certain linearization) result from different derivations.

In particular, the theoretical proposal defended in this paper is that the linearizations of declarative main clauses and non-embedded complement clauses that are observable at PF correspond to two types of discernible FinP-V2 constraints that need to be postulated for German:

▪

a discourse-related/information-structurally-motivated V2 (V2₁);

▪

a merely syntactic V2 (V2₂).

V2₁ is a generalized rule for (declarative) main clauses and involves a combination of V-to-Fin movement, (Fin° being the landing site of the finite verb) and cyclic movement of one constituent via Spec,FinP into some left-peripheral specifier (Spec,TopP, Spec,FocP, Spec,WhP, etc.) attracting it in order to satisfy some discourse or information-structural feature. In other words, V2₁ is construed as part of a strategy activating the entire C-domain to make room for (more or less overtly realized) phenomena associated with the presence of the speaker in the utterance. V2₂, instead, occurs in some types of V2 embedded clauses that do not allow for the realization of an independent speech act. In such structures, the linear order (in particular, raising of the finite verb + one XP into the C-domain) is not driven by any discourse or information-structural operation, but merely depends on the absence of an overt complementizer in C that forces these elements to move overtly. In V2₂-configurations, I assume the tree to be truncated at FinP, so that the finite verb moves to Fin° just as in V2₁, and the XP is not able to move into some CP-specifier after deriving the bottleneck/satisfying the EPP feature in FinP:

(31)	a.	[ForceP [TopP [XP]i … [FinP ti [Fin° Vfinx [TP ti … ty ]]]]]	(V21)
	b.	[ForceP [TopP [FocP/whP … [FinP [XP]i [Fin° Vfiny [TP ti … ty ]]]]]	(V22)

The distinction proposed here is advantageous for a number of reasons:

-

it provides a structural explanation of the (optional) presence of linear V3 in main, but not in embedded clauses complementing (i.e., in the scope of) non-assertive predicates. In combination with the bottleneck effect, it also accounts for the ungrammaticality of any Verb-Late(r) configuration that cannot be traced back to the derivational mechanisms spelled out above;

-

it is compatible with the identical interpretation of V2₂ and verb-final structures in embedded clauses (crucially, main-clause V2₁ cannot instead be realized by means of other linear orders).

Note that the idea of the existence of a ForceP-V2 type is dismissed with respect to the two types considered here, but not in toto, i.e., in relation to other V2 patterns of German not reviewed in the present paper or for other V2 languages. This study may hopefully pave the way for future cartographic investigations of the taxonomy of V2 in German and other languages.