Two structures of extraposition in central dialects of Early New High German

Extraposition in OV Germanic languages is a complicated phenomenon, which has been analyzed in different ways—as a result of rightward movement, as base generation of extraposing items to the right, and as a result of raising of predicative elements. The most common type of item in extraposition is CP, but also PPs and heavy adjuncts are attested in extraposition in Modern German. Early New High German (ENHG) material adds complexity to this topic, since it shows much more widespread extraposition, allowed for arguments as well. In this article, I propose a twofold analysis of extraposition in ENHG. Based on DP-extraposition to the predicate, I argue that in certain cases extraposition can result from rightward movement of an extraposing XP, while in other cases it is raising of the particle to position adjoined to vP, which leaves XP in overt extraposition to the predicate.


Introduction
This paper considers the phenomenon of argument extraposition in Early New High German (1350-1650 A.D.; ENHG in the remainder of this paper). Although ENHG extraposition has long been a focus of descriptive work, it has been under-studied in formal frameworks such as Minimalism. Such a study, however, can be useful both for a universal description of extraposition's nature and mechanisms, as well as for resolving the puzzle of German extraposition (see the discussion of different syntactic theories in Sect. 3).
& Aleksandra Belkind aleksandra.belkind@uni-leipzig.de In an SOV language, such as Modern and Early New High German, extraposition is easily detectable in subordinate clauses, as well as in main clauses with complex predicates. Every item to the right of the verbal complex 1 , i.e. in the postfield ('Nachfeldbesetzung' in the German tradition), can be defined as having been extraposed. The postfield is non-empty if there is any element following the right boundary of the clause, which is marked by the finite verb in subordinate clauses or the non-finite part of a verbal complex in main clauses. The term extraposition is used here and throughout the paper descriptively, referring to the surface position of an element in the postfield of a clause in any SOV language. Before the analysis is presented, I will remain agnostic about what structure(s) extraposition corresponds to. However, the aim of this paper is to argue that extraposition can result from two different structures.
In this study I concentrate on argument DP extraposition and demonstrate that in ENHG it corresponds to two different structures, namely (i) rightward movement of an extraposing phrase (proper extraposition) and (ii) raising of a small clause in the complement of VP to a higher projection adjoined to vP (SC-intraposition). (

1) a. [ CP …[ vP t i V]] X i b. [ CP V j … [[ SC (t k ) Y] i [ vP X k t i t j ]]]
Additionally I argue that the choice between these two constructions depends on both syntactic and prosodic factors: the structural characteristics of the nonextraposing part of the VP and prosodic phrasing of the extraposing XPs. The former factor ensures that only a complement of the V-head can raise under SCintraposition and hence that extraposition is separate from head-movement operations. The latter factor (the prosodic phrasing of the extraposing XPs) ensures the complete impossibility of multiple argument extraposition, but allows for argument+adjunct(s) extraposition via SC-intraposition. In what follows I will systematically consider Early New High German argument extraposition and provide arguments for each of my claims. If it is true that the choice between the two extraposition constructions depends on the structure and prosodic weight of the right boundary and prosodic phrasing in the clause, then arguments for different theories of extraposition in Modern German and other languages can be unified.
The rest of this paper will proceed as follows: In Sect. 2, the design of the study and some crucial characteristics of the ENHG argument extraposition are considered. Section 3 discusses existing theories of extraposition. In Sect. 4, I propose two mechanisms of extraposition and discuss their precise mechanics, structural constraints and predictions. Section 5 deals with empirical and theoretical consequences of the prosodic nature of extraposition. Appendix briefly discusses the extraposition of adjuncts, and Sect. 6 concludes. Sapp (2014) argues that genre has more impact on extraposition rate than dialect. The Thuringian texts are examples of technical prose and therefore differ stylistically and pragmatically from the others and are expected to favor extraposition more than the rest. Though Sapp also argues that religious texts and chronicles have equally low extraposition rates, all chronicles in my sample have a good amount of extraposition, but the second Ripuarian text, 'Vonn warer, wesenlicher, vnd pleibēder Gegenwertigkeit des Leybs und Blůts Christi ...' of Johann Gropper, is almost irrelevant for the analysis because of a complete absence of extraposition. Nevertheless, I follow Sapp (2014) in attributing this difference not to the dialect, but rather to stylistic properties of the text. For this reason, I also argue that the absence of extraposition in this text does not provide counterarguments against my results, nor does it falsify them. 2 Corpus output was collected for ten out of the seventeen particles which comprise the core of particle verb combinations (Habermann 2011). All the most frequent particles were considered, as well as some less frequent ones. Verbal particles are not glossed in the Bonner Frühneuhochdeutsch-Korpus either as separate items, or together with verbs. For this reason, automatic search in the ANNIS system was based on a token sequence for each verbal particle. Contexts with PVCs were manually selected from the search output, which included synonymous prepositions and identical token sequences in the beginning of other words, e.g., 'an-' in 'ander'. Because of the search design, I provide a list of the particles represented in the corpus

Data description
The most common types of extraposition in ENHG are CP-and TP-extraposition (Sapp 2014). However, these are also the most common, or even obligatory, types in Modern German and thus cannot add much to the discussion of extraposition. In contrast, DP-extraposition is almost impossible or at least highly limited in Modern German. In ENHG, on the contrary, extraposing DPs are quite common even in the written language. Given that ENHG is very similar to Modern German in its main properties, differences between the two languages are interesting per se and can potentially be useful for an analysis of extraposition in general. The amounts of postposed DPs in my data are distributed as follows: in the first period (1450-1500) there are 5 cases in Thuringian, 20 in Hessian, and 17 in Ripuarian out of the overall number of 245 clauses with extraposition. In the second period (1550-1600), however, the numbers are smaller: in Thuringian, 4 cases; in Hessian, 3; and in Ripuarian, 0, out of a total number of 297 clauses with extraposition. These numbers show a drastic reduction in argument extraposition over timethere are 42 cases of DP/NP extraposition in the first period (1450-1500 A.D.), but only 7 cases in the second period (1550-1600 A.D.). A similar reduction in frequency can be observed for extraposed PPs and adverbials: compared to 39 cases in the first period, there are only 19 in the second. As for extraposition of CPs and infinitival clauses, there is an opposite tendency: 19 cases in the first period and 73 in the second one. Despite the frequency reduction, both the DP/NP and PP/Adv extraposition types are present in both periods. This difference in frequency may be attributed to diachronic change. Nevertheless, even in the second period, the number of DPextraposition instances is sufficient to show that this type of extraposition is present in the language. Equally important is the fact that the data does not show any radical For the purposes of the present study, I do not consider the most permissive judgments, because they are not accepted by all native speakers and lack a prior analysis. However, I do not make any strong arguments about Modern German because the intuitions of native speakers differ and, as pointed out above, a separate corpus-and statistics-based study is required.
The same is true for extraposition of direct objects: ENHG allows both heavy and light direct objects in extraposition, while Modern Standard German does not. Truckenbrodt (2016) argues that direct object extraposition in only possible in Modern Standard German if the direct object is heavy, as in (4). 6 Early New High German, however, also allows extraposition of light direct objects (5). These contrasts show that extraposition in Early New High German was far more widespread than in Modern Standard German. However, it was not unrestricted. Given that there is no possible access to negative judgments for ENHG and that both subject and object extraposition is attested quite well in my data, I assume that the absence of a particular extraposition configuration suggests that it is unavailable in the written ENHG. 7 Examples in (6) provide an illustration of such unattested extraposition configurations:  6 As an anonymous reviewer noted, some native speakers do not consider this sentence grammatical with either heavy or light DO. In the ENHG, on the other hand, this is a common pattern. 7 As another anonymous reviewer points out, there can be a difference between spoken and written language. Since we have no access to the spoken variety of ENHG, any claim in this paper holds only for written ENHG. I would like to emphasize, however, that the same is true for any other study of ENHG. Hence, ENHG stands in fact for the written variety of German dialects, spoken in 1350-1650 A.D. The examples in (6a-c) illustrate unattested patterns of extraposition in clauses with a participial and a finite right boundary. The same holds for all clause-final verbs, whether they are infinitives, participles, or clause-final finite forms. If these restrictions hold indeed, a theory of extraposition should be able to account for them.
Before proceeding to the analysis, I will review the existing theories of extraposition.

Theories of extraposition
There are three main groups of approaches, which treat extraposed elements either as moved to the right or as base-generated, either in their canonical position, or to the right of V. Truckenbrodt (2016) considers extraposition in Modern German to consist of two different phenomena: (i) proper extraposition, i.e. an element belonging to the clausal structure (7), 8 and (ii) right-dislocation/afterthought, i.e. the extraposing item belongs to neither the argument structure nor the prosodic structure of the clause (8). Right-dislocation and afterthought are rather irrelevant for the present study since they do not belong to the syntactic structure of the preceding clause. In what follows, I concentrate on analyses of proper extraposition.
In the literature, there is no consensus on either the nature of extraposition, or on what elements can extrapose. It is agreed upon that there is CP-extraposition, which is possible for both CP-arguments and relative clauses. 10 Less widespread, but generally allowed, is PP-extraposition. (e.g., Haider 1997;Müller 1997;Büring and Hartmann 1997;Wurmbrand and Bobaljik 2005;Frey 2015). DP-/NP-extraposition is generally ignored in the syntactic literature, despite the fact that it is at least marginally attested in Modern German, as examples from Truckenbrodt (2016) and a brief consultation with German native speakers show. As for the nature of extraposition, present theories fall into three main groups that consider extraposition to be either (i) a result of base-generation of a constituent to the right, (ii) a result of base-generating a constituent in its canonical place, followed by raising of V, or (iii) an instance of rightward movement of the extraposed material.
Base-generation theories of extraposition (Culicover and Rochemont 1990;Wiltschko 1994) suggest that extraposed clauses are base-generated in rightadjoined positions. However, as soon as argument extraposition is considered, this analysis becomes problematic.
An alternative base-generation approach is based on Kayne's (1994) Antisymmetry Hypothesis. If the universal order 'specifier-head-complement' is assumed, the OV order emerges by movement of the direct object to a higher position. Extraposition is then analyzed as an absence of raising, keeping the original SVOorder. 11 Along with other problems sketched below, this analysis would not be able to account for extraposition of external arguments, since these should be first merged in Spec,vP (e.g., Harley 2017) and thus do not follow the verb. In the next section, I will show that external arguments can extrapose in ENHG. Haider (1997) suggests an analysis where an extraposed constituent is basegenerated in a left-branching OV structure, followed by raising of the verbal material out of VP. His core argumentation is based on the CP position relative to the position of the standard of comparison in comparative constructions. Büring and Hartmann (1997) point out a number of problems that basegeneration approaches induce. As examples (9-10) show, some but not all extraposed CPs are islands for extraction: (9) [= 9 Büring and Hartmann (1997) However, if all extraposing CPs are base-generated in Comp,VP and do not move, no island effects should emerge. Under a rightward movement analysis, the difference between subjects and objects is determined by their base position because extraction takes place from there, prior to extraposition.
A second problem for the base-generation approach is Haider's 'Binding paradox'. Examples in (11) show that an R-expression in an extraposing relative clause can bind a pronominal indirect object. If the R-expression is in an object clause, it cannot be coreferent with a higher pronoun.
(11) [= 19a,b Büring and Hartmann (1997) Under the base-generation analysis this is indeed a paradox since both CPs are expected to be generated in the same position. But the movement analysis can account for this pattern if the relative clause is first merged above the indirect object and thus c-commands it, while the object clause never does. Büring and Hartmann (1997) show that the base-generation account is not only too permissive, but also too restrictive. As one example, an extraposition of a relative clause out of a PP is allowed (12). This is only possible if the relative clause is base-generated as a complement of the NP 'Professorin', because otherwise N cannot c-command the relative clause and bind the relative pronoun. On the other hand, Haider (1997) points out some complications for the rightward movement theory. The first problem is the absence of island effects in, e.g., object-CPs (recall ex. 10a). However, Müller (1997) suggests that this can be accounted for if (CP-)extraposition is analyzed as remnant movement. If an item is moved out of the CP before the latter moves to the right, no island effect is expected.
The second problem is the impossibility of topicalization and scrambling of object-CPs, which is unexpected if they can be A'-moved, i.e. extraposed (13).
Thirdly, Haider shows that focus particles can be topicalized together with the phrases they introduce, but cannot be extraposed. This difference is not expected if both operations are A'-movement.
However, at least the last complication can be resolved. It has been argued that extraposition is a PF phenomenon, tightly linked to prosody and information structure (e.g., Sapp 2014;Truckenbrodt 1995;Göbbel 2007;Wurmbrand and Bobaljik 2005;Belkind 2021, Ms.). Truckenbrodt (1995), and more recently Wurmbrand and Bobaljik (2005), argue for a treatment of extraposition as a choice of copy at PF: (13) Choice of copy in an extraposition chain: Pronounce the higher copy ... unless doing so interrupts the maximal parsing of the remaining material into a prosodic phrase.
Thus, the choice of copy is regulated by prosodic phrasing rules and does not need any separate syntactic motivation. However, this analysis, being a basegeneration one, runs into the same problems as described above (after Büring and Hartmann 1997). Nevetheless, it seems that prosodic structure is indeed the best motivation for extraposition (e.g., Belkind 2021, Ms.). Since prosody is tightly connected to information structure (Samek-Lodovici 2015Szendrői 2017; etc.), extraposition of focus particles can be seen as overgeneration of overt focus marking on one element.
As this section shows, there is no consensus about what extraposition is. Since there are strong arguments for treating it as a PF phenomenon, it would be an easy and logical solution to argue that it is an operation on linear order of elements that does not care about syntax. However, such an approach would be too permissive and would not straightforwardly predict any structural constraints on extraposition. Adopting Belkind's proposal that extraposition in ENHG, as a prosodically motivated reordering of elements, takes place postsyntactically, I will show that it nevertheless does not ignore syntax completely.

Analysis of ENHG DP-extraposition
In this section, I will argue that DP-extraposition in ENHG is best described as two different operations: either small clause raising over an extraposing XP or rightward movement of the extraposing XP. The choice of operation is sensitive to (i) whether the verbal complex consists of only one V-head or the matrix verb takes a small clause (SC) complement which forms a part of the verbal complex (e.g., verbal particle), and (ii) overt vs. silent copies of the verb in the VP.
This section proceeds as follows: First, I provide motivation for using particleverb combinations (PVCs) as the main piece of data for the analysis of extraposition. Subsection 4.2 contains arguments for the small clause raising analysis as well as structural constraints on this operation. In Subsect. 4.3, a rightward movement analysis is suggested for the rest of the extraposition contexts. In Sect. 5, constraints for each of the operations are postulated and a unifying analysis is proposed that makes it possible to treat both operations as extraposition, despite the different machinery used.

Structure(s) of PVC
Before proceeding to the analysis of ENHG extraposition, a few remarks on the structure of PVC should be made. PVCs provide good material for research. On the one hand, the particle (in the unmarked case) stays in situ and marks the right boundary of the clause, which allows one to test extraposition not only in perfect or passive contexts, but also in the present or preterite, when the verb raises to a higher position. (ENHG is not obligatorily V2 in main clauses, but the finite verb normally does not stay in situ (Schmidt 2013;Ebert et al. 2013).) On the other hand, PVCs form complex tenses and passive forms, as well as embed under modal verbs. Therefore, PVCs make it possible to detect extraposition both in clauses with a raised lexical verb and in clauses where the lexical verb remains clause-final (e.g., perfect, infinitives selected by modal verbs or causatives, subordinate clauses). As will be discussed in the next subsection, extraposition that follows a sole verbal particle differs in its properties from extraposition that follows a full verb.
There exist two main approaches to analyze the inner structure of PVC: the small clause approach and the complex head approach (e.g., Neeleman 2002;Müller 2002;Dehé 2015; see Wurmbrand 2000 for some further references). Vikner (2013) tries to reconcile these two approaches. He argues that particles form a syntactic phrase the same way as prepositions do, with the only difference that particles do not assign case (Vikner 2013, 2). There are two possibilities of how a DP can be assigned case: either the particle moves to the V-head and pseudoincorporates into the verb, or the DP moves to Spec, PtclP, where case can be assigned by V°. Though both options are possible, Vikner argues that particles are incorporated into the V-head. Wurmbrand (2000) suggests another way to combine the small clause and complex head analyses, arguing that particle verbs can have either structure, depending on their semantic compositionality. Semantically transparent PVCs have the small clause structure, while idiomatic (and semi-idiomatic) PVCs have the complex head structure.
(14) a. small clause structure b. complex head structure 12 In the remainder of this paper, I will adopt this approach as a baseline for my argumentation. I will also refer to these structures as transparent vs. (semi-) idiomatic particles, respectively. However, given the absence of negative judgements, as well as incomplete positive data, it is not possible to run all necessary tests. Therefore, all statements about differences in syntactic behavior of transparent vs. (semi-)idiomatic PVCs are to a certain degree speculative and should be taken with caution.

Small clause raising
In this subsection, I argue that some instances of extraposition can be only derived by raising of a small clause, which contains a non-silent part of a verbal complex. The proposed derivation is presented in (15). It takes place in main clauses with transparent PVCs and raised finite verbs.
The derivation in (15) has several consequences. First, small clause raising will only lead to extraposition if the matrix verb raises from its base position. Secondly, particle raising is an instance of phrasal movement. In addition, it requires the direct object to move out of the small clause beforehand. Thirdly, the landing position of the raised small clause is an (adjoined) projection above vP. In what follows, each of these predictions is shown to be borne out. In Subsect. 4.2.4, multiple extraposition is introduced as evidence against a rightward movement analysis of these clauses.

Verb raising out of vP feeds particle raising
In this subsection, I show that extraposition can be derived via small clause raising only if the lexical verb is finite and moves out of the vP.
Examples in (16-17) show that multiple extraposition of an object and adjunct(s) is only possible if the lexical verb is finite and raised to TP (or to CP). Crucially, no example like (17b) was found in the corpus. A possible complication for the small-clause raising analysis can arise from the fact that in (16) the direct object and the adverbial seem to be additionally scrambled. Such a word order is expected if both phrases are right-merged in their base positions. However, a right-adjunction analysis would not be able to rule out (17b). A simple solution for obligatory scrambling in extraposition is provided in Vikner's (2013) proposal that objects are first merged as complements of a particle but assigned case in the Spec, VP. However, it should be noted that the second part of Vikner's proposal, i.e. subsequent particle incorporation into the V-head, is impossible in the analysis presented here since it would require double movement of a V-head-the verb raising to the V2 position and the particle raising to an adjoined projection above vP.
Hence, I propose the derivation in (18): First, the direct object moves above the adverbial to Spec, VP in order to get Case; second, the SC undergoes remnant movement to Spec, vP, leaving both the direct object and the adverbial phrase in extraposition; and at last, the verb is raised to C.
This derivation makes an additional prediction. If a verbal particle can move to a higher position in extraposition clauses, this operation should also be available when the main verb stays inside VP. This is indeed the case, as is shown in the next subsection in (19).

Phrasal movement
In the derivation in (18) the particle moves with the small clause, and not through head movement. This proposal consists of two parts: (i) the particle is the head of a small clause and not incorporated into the verbal head, and (ii) the particle moves together with the whole small clause, and not as a head. Additionally, (iii) if an object is introduced in the small clause, it should be raised together with the small clause, unless it escapes the small clause before it moves. In what follows, I address each subclaim in turn.
As has been already mentioned above, I adopt a small clause analysis for semantically transparent PVCs (e.g., Wurmbrand 2000). Under this analysis, a verbal particle is not incorporated into the V-head (contra, e.g., Vikner 2013) and is expected to be able to move independently from the verb. This is borne out, since transparent particles in ENHG can be raised while the lexical verb stays in situ: The word order in (19) seems to be the result of a separate movement of the particle, with the PP-object do mit not having undergone the raising to a higher position. With do mit in situ, the raised particle is separated from the verb. For this to be possible, the particle must not be incorporated into the V-head. With respect to extraposition, this means that raising of a particle can be either head movement or phrasal movement of the whole small clause. I argue that the latter is the case. The first argument for particle raising as (remnant) phrasal movement comes from causative constructions. In (20) (20) looks like a counterexample to the proposal that multiple extraposition past overt copies of verbs is banned. However, there is a crucial difference between an infinitive in a causative construction (as in (20)) and, for instance, an infinitive in an infinitival clause: the infinitive in causative construction is the head of a small clause. Pitteroff and Campanini (2013) argue convincingly that causative verbs in Modern German such as lassen 'let' are phase-selecting; i.e., they take VoiceP 13 as their complement (21). 14 Thus, causative constructions have a structure very similar to that of transparent verbal particles (repeated in 22) and allow phrasal movement. 13 In the current paper, I do not split the Voice domain into Voice and v, but assume a simple bundled vP.
14 Though different causatives do take complements of different size (Pylkkänen 2008;Alexiadou et al. 2015), I do expand this analysis to ENHG machen-clausatives. Speyer (2018) shows that in MHG and ENHG accusativus cum infinitivo (ACI) had in general a uniclausal structure. Though he does not analyze causative constructions, the same analysis can be applied to them as well, considering the analysis of Pitteroff and Campanini (2013) for Modern German.
A second argument for phrasal movement comes from the inability of idiomatic particles to raise. Recall Wurmbrand's (2000) hypothesis that semantically transparent verbal particles differ from semantically opaque ones (idiomatic and semi-idiomatic) in that the former have a small-clause structure (22) while the latter form a complex head with a verb (repeated in 23). 15 (23) An immediate consequence is that small-clause raising is unavailable since the particle is now in the V-head. Hence, idiomatic PVCs do not allow extraposition via SC-raising. This prediction seems to be borne out. Though it is impossible to run syntactic tests (e.g., those in Wurmbrand 2000) in order to diagnose a PVC to be a transparent or (semi-)idiomatic one, and the grammaticality of the ENHG data cannot be judged based on the intuitions of speakers of Modern German, I was not able to find any examples of multiple XPs (argument DP + adjuncts) linearized to the right of a clearly (semi-)idiomatic particle in my corpus data.
There is an additional derivational step, predicted by the phrasal movement proposal: if a particle moves together with the small clause, it is remnant movement. In other words, the object should escape from its base-position in the complement of the small clause beforehand, in order to be linearized to the right. I argue that the object obligatorily raises from its base position in all clauses, irrespective of whether remnant movement of the small clause takes place. As Vikner (2013) points out, particles cannot assign case. Hence, objects base generated as complements in the particle phrase cannot receive case in their base position. He suggests two ways out: (i) the DP moves to the Specifier position of the small clause where it can get case from the verb in an ECM-like configuration (24a); (ii) the particle gets incorporated into the verb and this complex head assigns case to the DP (24b). Vikner argues for the second option. However, as discussed above, the incorporation approach cannot account for a particle raising separately from the verb. On the contrary, DP-movement not only creates the environment for phrasal movement of the SC, but also independently explains how the object escapes from the moved phrase.
(24) a. DP-raising analysis b. particle incorporation analysis An additional argument for the DP-raising analysis comes from the relative order of the object DP and adjuncts. As already pointed out in Sect. 4.2.1, extraposed objects universally precede low adjuncts. This follows naturally if DP moves to a position above the modifier projection.
However, the derivation in (24a) has two disadvantages. First, DP-raising is triggered by case assignment. Hence, it either happens before V is merged and is thus unmotivated at this step of the derivation, or it violates the Extension Condition. Secondly, a movement step from the Comp, SC to the Spec, SC is too short and violates Antilocality (Abels 2003;Deal 2019). Therefore, I suggest Spec, VP as the landing site. In causative constructions, the causee is introduced in the Spec, CauseP and gets case-licensed there (e.g., Pylkkänen 2008). Similarly, if the direct object moves from Comp, SC to Spec, VP, it does not violate antilocality and can get case (25).

(25)
To sum up, the main arguments for phrasal remnant movement are: a. Phrasal movement explains why there is a difference between transparent and (semi-)idiomatic particles, i.e. why particle movement is only available for transparent ones. b. It also explains why multiple extraposition is possible after an embedded verb in a causative construction, but there is no such option for participles in complex tenses. c. For phrasal movement to be allowed, the direct object should first move out of the SC. This prediction is borne out. Objects in extraposition land higher than adverbials modifying the VP/SC (see examples 16 and 20 above). Additionally, Object movement to Spec,VP might be independently required for case assignment.
Analyzing these instances of extraposition as a result of remnant movement has an additional advantage. Müller (1997) argues for CP-extraposition in Modern German to be remnant rightward movement. If the particle or the embedded vP movement is also remnant movement, both types of extraposition show similarities -an advantage for attempts of a unified analysis. In Sect. 4.3, I will provide arguments for rightward movement to be present in the ENHG argument extraposition, as well, but for now I will leave this topic and turn to the landing site of the raised SC.

Landing position
There are three possible positions for a particle to land lower than the finite verb in C°(or T°): • vP-adjoined projection, i.e. the lowest position available to enable subject extraposition; • Some functional projection above vP, available due to some adverbial and aspectual semantics of transparent particles; • T domain, where the particle gets pied-piped with the verb, or independently raises to Spec, TP.
The projection adjoined to vP is the most uncontroversial option and is supported by extraposition in causative constructions.
Particle movement to the T-domain is a possible option, but leads to severe problems. There are two available landing sites in the T-domain-T°, i.e. piedpiping with the verb, and Spec, TP-but both of them are incompatible with either the theory or the data. T°is an unsatisfactory option for particle movement for various reasons. First of all, it should be occupied by a silent copy of the finite verb. Secondly, T°bears a tense feature that cannot be interpreted on the particle. Thirdly, if particle raising is indeed an instance of phrasal movement, it cannot be pied-piped to T°and then stranded when the verb is moved higher to C°. As for Spec, TP, it seems to be less problematic theoretically, but various material can occur between a finite verb and a raised particle. These examples allow to conclude that the landing site for a moved particle in the T-domain is rather unlikely.
An adverbial projection is not an obvious option for movement of the head of a small clause. However, it is well known that, etymologically, verbal particles used to be locative adverbials (e.g., Grimm, Bd. 2 1826, 698). Hence, small clauses built by transparent particles can be seen as headed by an adverbial-like item and able to move to an adverbial projection higher in the tree. Alternatively, particle movement to AspP can be proposed (e.g., Svenonius 1996;Grewendorf and Poletto 2012;Quaglia and Trotzke 2017). However, this analysis is not borne out. In the previous subsection, it was shown that multiple extraposition to vP in causative construction is possible (repeated in 27). It is unlikely that particle raising and VP-raising in (27) have different landing sites since lower adverbs occur to the right of the moved infinitive (27a), and higher sentential adverbs occur to the left (27b). However, a VP lacks features which would trigger movement to an adverbial projection or to AspP. Hence, the only available landing position for a small clause or infinitival phrase is a projection above the vP. Since I believe this type of movement to not be feature driven, I assume that it is adjunction, rather than raising to a functional projection.

Evidence against rightward movement: multiple extraposition
One of the most prominent theories of extraposition and an alternative to the proposed analysis is a rightward movement theory. However, multiple extraposition provides an argument for the small clause raising analysis.
As examples in (28) and (29) show, multiple extraposition of an argument and adjunct(s) is possible for both subjects and objects. Simultaneous extraposition of several arguments is absent in my sample, which I take as evidence for it to be ruled out. However, as will be discussed in Sect. 5, it is ruled out due to prosodic phrasing rules, and not due to syntactic reasons. Examples (28-29) show an important asymmetry in the relative position of adverbials and arguments: in (28), the subject follows subject-oriented adverb 'williglichen', while in (29), the objects precede place adverbials. This is true for all examples of multiple extraposition in the corpus. On the one hand, it shows that object DPs in extraposition should be moved out of their base position. On the other hand, in (28), the subject and the adverbial keep their base-generated ordering: the adverbial 'williglichen' ('willingly') is subject-oriented and can be merged only if Spec,vP is already occupied by an external argument (e.g., Pylkkänen 2008;Alexiadou et al. 2015;Harley 2017). Therefore, it is merged after the external argument is introduced and is higher than the argument DP. To sum up, adverbial +subject extraposition keeps the base-generated order, while object+adverbial extraposition shows a reversed order of elements. If both XPs are moved to the right, this asymmetry is accidental. In contrast, under the small-clause raising analysis, it emerges naturally.
Based on this, I assume small-clause raising to be the optimal analysis. The derivations for subject and object extraposition are presented in (30) 16 and (31).
A further argument against the rightward movement analysis comes from the motivation for extraposition. Under the rightward movement analysis, each phrase moves separately. Hence, each one should have a separate motivation. However, this is hard to model: since all candidates for extraposition are in the same Spell-Out domain, the order in which they should be moved to the right does not emerge naturally. There is no way to identify an order in which the same movement operation should be applied to several elements with more or less similar triggers for this operation. 17 Thus, the derivation would crash because it is impossible to define the order of several identical operations.
In this subsection, I have introduced arguments for small-clause raising and against the rightward-movement analysis of multiple extraposition in ENHG. These are the relative position of arguments and adverbials and problems with modelling the timing of multiple identical movement operations in the same Spell-Out domain.

Rightward movement
This section aims to show that not all cases of the ENHG extraposition can be analyzed as a result of SC-raising. Rightward movement proposed for CPextraposition in Modern German (e.g., Büring and Hartmann 1997;Müller 1997) also occurs in ENHG when SC-raising cannot take place.
As was pointed out in Sect. Nevertheless, this is the sole example of this sort I have been able to find. More importantly, there are at least two alternative analyses which are consistent with the present theory and which cannot be ruled out. First, 'troest in den tzijden der bedroeffnisse' can be one constituent, where the PP is either a modifier or a complement of the noun. The second option is to treat the PP as an afterthought, which is possible because of its position at the edge of the sentence. In order to rule out each of these analyses, access to prosodic information is required. If the PP is not a part of the DP, it should form a separate prosodic phrase. As an afterthought, it should have a specific prosodic marking. Since neither of these analyses can be tested, I do not take the example in (i) as a counterargument to my proposal. Out of these patterns, (32b) is the most important one. Object extraposition is well represented in the corpus (32 cases), as well as object + adjunct(s) extraposition (7 cases). However, an object + adjunct(s) string can only follow a transparent particle, but not a verbal stem. I take this as evidence that this structure is unavailable in the language. A model of extraposition for ENHG should be able to predict this restriction. In the remainder of this subsection, I present counterarguments against a small-clause raising analysis of extraposition after verbal forms. As a consequence, it is necessary to assume the existence of another operation leading to the extraposition structure. A good candidate is the rightward movement analysis. It is not only able to capture the data, but was already argued for in studies about Modern German extraposition.
The restriction on multiple extraposition after full verbal forms cannot be derived in the raising analysis. On the contrary, the SC-raising analysis was proposed for ENHG in order to capture multiple extraposition. As the examples above show, multiple extraposition is unavailable if a verb stays clause final.
The simplest way to save the SC-raising analysis is a VP-raising to the same position. However, VP raising to Spec,vP would violate antilocality as Comp-to-Spec movement (e.g., Abels 2003). Moreover, object movement to Spec, VP would be necessary because, otherwise, it would be pied-piped with the rest of the VP. This also violates antilocality. Object movement to Spec, vP would solve the antilocality problem, but this lacks any independent motivation since it is absent in non-extraposition clauses. An abstract derivation in (33) gives a summary of the problems just listed.
(33) Yet another argument against the VP-raising analysis comes from the position of low adverbials before the verb (34). If the VP is raised, it should be raised above the low adverb, similar to SC. But this is not reflected in the surface word order. In (35)(36), a participial manner adjunct shows up in the extraposition domain or outside of it, depending on the position of the verb: In the previous section, I have argued that multiple extraposition of arguments and low adverbials follows directly if a SC raises to a position above the vP and everything inside the vP automatically occurs in extraposition. In (35), low adverbials are linearized before the verb. This, taken together with antilocality problems, rules out a VP-raising analysis. Hence, there are two logical possibilities to derive a word order as in (35): (i) base-generation of an argument in a rightbranching complementizer/specifier projection, and (ii) rightward movement of a DP. The second option has an advantage since it allows the internal argument in (36) to be passivized and rise to a position where it can get nominative case (Harley 2017 Based on these arguments, I assume (i) that the VP-raising analysis is not supported, despite its advantage being very similar to the SC-raising analysis, and (ii) that rightward movement of the DP-extraposition to full verb forms is most preferable since it captures all relevant data and allows to combine passivization with extraposition.

Interim summary
In this section, I have shown that extraposition in ENHG is best analyzed as two distinct operations-small clause raising and rightward movement. The choice of operation is dependent on the structure of VP (i.e., presence vs. absence of a small clause in the complement of VP) and on the position of the lexical verb (raised out of vP vs. in situ). In the next section, I will address the question why these two derivations can still be treated as one phenomenon, and what constraints on extraposition exist.

Prosodic nature of extraposition
This section argues that both operations described above can still be unified under a notion of extraposition. Despite technical differences, they share a prosodic motivation and, crucially, a constraint on multiple argument extraposition which can be only motivated in prosodic terms.
Multiple extraposition of an argument and one or several adjuncts is allowed when it results from SC-raising to a projection adjoined to vP. There are, however, two constraints: multiple rightward movement, as well as multiple argument extraposition in general are not attested in my data. I argue that this is not a coincidence, but a motivated pattern.
(38) Constraint on multiple argument extraposition: If an argument X is extraposed and follows a verbal particle, a verb cluster or a verb, no other argument Y is allowed to extrapose.
These constraints are both prosodic in nature. As argued by Szendrői (2017), extraposition forms a separate prosodic phrase. Crucially, arguments and adjuncts differ in the mechanism of prosodic phrasing: arguments in the middle field are included in a bigger prosodic phrase (=VP). Adjuncts, on the contrary, form separate prosodic phrases, irrespective of their position in a clause (Samek-Lodovici 2005). Hence, adjuncts do not change their prosodic phrasing characteristics from the middle field to extraposition. This is not true for arguments. Arguments in the middle field all belong to one and the same prosodic phrase, but in extraposition, they form each a separate prosodic phrase. This captures the difference between extraposition derived with SC-raising and rightward movement. SC-raising leaves the whole vP in extraposition. This leads to the formation of a separate prosodic phrase of an argument XP. But adjuncts remain unaffected. Hence, only an argument XP needs motivation for prosodic rephrasing. 19 Under rightward movement, however, each XP moves separately, and therefore, each requires separate motivation. Since the most common motivation for extraposition is contrastive focus (Sapp 2014), multiple rightward movement would lead to two emphasized phrases in a row. This should be ruled out as a prosodically unlikely configuration. However, Szendrői (2017) argues that defocusing can trigger extraposition, as well. If one phrase is contrastive and the other one defocused, multiple rightward movement would not lead to two emphasized phrases in a row. Nevertheless, this configuration can be ruled out, as well. If extraposition takes place during Spell-Out, two phrases will be able to trigger rightward movement in the same Spell-Out domain. Since motivations for extraposition are not hierarchically ordered, it is impossible to define the order in which both phrases should undergo rightward movement, and the derivation crashes. Hence, only one instance of rightward movement can be allowed for each Spell-Out domain.
In this section, I have briefly shown that two extraposition configurations are ruled out by prosodic phrasing rules. Since extraposition is argued to be derived during Spell-Out, only one operation is allowed per Spell-out domain. This presupposes that only one XP can trigger extraposition; otherwise, a conflict of interests emerges, and the derivation crashes. Based on that, both multiple rightward movement and multiple argument extraposition can be easily ruled out as configurations requiring several XPs that have independent-and conflictingmotivations for extraposition. 19 This also explains why extraposition of a subject XP is only possible if it is the sole argument of the verb. Anagnostopoulou (2001, 2008) argue that due to the Subject-in-situ Generalization, only one NP-argument can stay vP internal, while all others have to be evacuated. This would explain constraint (38). Note that this generalization does not quite work in Modern German, where multiple arguments inside vP are allowed (Wurmbrand 2004;Alexiadou and Anagnostopoulou 2008). However, as pointed out before, due to the prosodic nature of extraposition, only one argument has to occur in the postfield. Hence it is crucial that the in situ subject would be the only argument, when in extraposition.

Conclusion
Extraposition has been a long-standing problem in linguistics. With this study, I aim to contribute to the discussion about the syntax of extraposition in Germanic SOV languages by providing an analysis of argument DP extraposition in Early New High German. Despite the complications unavoidable in an analysis of a nonmodern language, Early New High German can offer a relevant piece of data to the general discussion since argument extraposition seems to be absent-or at least highly restricted-in modern German and Dutch. For ENHG, on the contrary, DPextraposition is a widespread phenomenon.
The analysis proposed in this study suggests that DP-extraposition in ENHG cannot be analyzed as one movement operation. Instead, DP-extraposition results from two distinct operations, where the choice of operation depends on the inner structure of the VP. The first type of derivation takes place in main clauses with a lexical verb raised out of its base position in vP. If, after verb raising, the right boundary of the clause is still marked with an embedded vP or a small clause headed by a transparent verbal particle, the embedded phrase moves to the left edge of the Spell-Out domain. Since such an operation will not bring about the desired result, if the V-head is occupied by an overt copy of the main verb, a second type of derivation is introduced. In this second type of derivation, extraposition is the result of rightward movement of a DP.
Each of the proposed derivations has certain consequences. First, the raising operation should be an instance of phrasal movement. Secondly, it feeds multiple extraposition of a DP and low adverbs. Thirdly, it presupposes that under object extraposition, an object DP should first move out of the small clause. This movement predicts that the object should precede adjunct modifiers of the small clause. The subject, however, stays in situ. Hence, it should follow a subjectoriented adverbial modifier of vP. If, on the other hand, the raising derivation is not available and DP extraposition results from rightward movement, no multiple XPs can occur in extraposition. All these predictions are borne out.
Despite having two derivations, I argue that extraposition is still a coherent phenomenon since it has one motivation. Moreover, it is not a purely syntactic phenomenon because it is triggered by prosodic phrasing, which should be inaccessible in syntax. During Spell-Out, however, all three necessary components are available: (i) the hierarchical structure of the phase is not yet flattened, (ii) information about head movement should be accessible, and (iii) since Spell-Out happens at PF, prosodic information can also be assumed to be active.
Funding Open Access funding enabled and organized by Projekt DEAL.
Availability of data and materials All data are publicly available in Bonner Frühneuhochdeutsch-Korpus.

Declarations
Competing interests The author declare that they have no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Thus, even in cases which require arguments to move rightward in order to extrapose, multiple extraposition of adjuncts is possible. There are three logically possible ways to model this: a. VP-raising to Spec,vP b. Multiple rightward movement of each extraposing adjunct c. Base-generation to the right.
The problem with the first model is that it violates the antilocality restriction on Comp-to-Spec movement (see tree-derivation in (2)). (2) The next model, namely, multiple rightward movement, would be in principle possible, but only if the constraint on multiple rightward movement (presented in (37) in Sect. 5) were not active. Since this constraint ensures that there is no multiple extraposition of arguments + adjunct(s) to a non-silent V-head, such a solution would be unsatisfactory.
Finally, the last option-base-generation of adjuncts to the right-seems to be a plausible way out. Frey (2015) suggests this model for Modern German. By analyzing c-command relations and Principle C violations, he argues that adverbials and PP-arguments are base generated in extraposition, while attributive subordinate clauses are moved to the right post-syntactically. The structure of the postfield ('Nachfeld') is identical to the right-branching structure of VP (initially proposed in Larson 1988). Adverbials, on the other hand, are base generated in the same positions as in the middle field with the only difference that, when postposing, they right-adjoin and appear in the reversed order (3).
(3) a. Sentential adverbials \ mood adverbials \ Subject \ time and place adverbials \ Objects \ manner adverbials \ verb b. Sentential adverbials \ mood adverbials \ Subject \ verb \ objects \ manner adverbials \ time and place adverbials Crucially, Frey's theory does not exclude additional scrambling of the adjuncts in extraposition, which makes it impossible to test whether this theory captures ENHG data, e.g. whether the order of the adverbials in (1a) and (1c) is a result of scrambling or base generation.
As pointed out in Sect. 5, adjuncts form separate prosodic phrases irrespective of their position in the clause. Scrambling results in a specific stress pattern, but when there are only written sources available, no information of this sort can be found. Therefore, adjunct extraposition cannot be used as an argument either for or against the model of extraposition in ENHG presented in this section.