4.1 A subject requirement in V3

A nominative subject requirement in V3 is attractive because preverbal accusative object DPs are largely absent in the data (Wiese & Rehbein Reference Wiese and Rehbein2016, Walkden Reference Walkden2017) and reportedly rejected by speakers (Wiese, p.c. in Walkden Reference Walkden2017:55), although Schalowski (Reference Schalowski2017) reports one exceptional instance (6).

Multiple attestations of preverbal locative (7a,b) and temporal adverbs and adverbial DPs (c– e) (cf. Schalowski Reference Schalowski2017, Walkden Reference Walkden2017) challenge a strict preverbal subject requirement (contra te Velde Reference te Velde2017).

While spatio-temporal adverbs can behave like DP subjects, as argued (controversially) for Locative Inversion in English (Bresnan Reference Bresnan1994, Hartmann Reference Hartmann2008), such scenarios rely on unaccusativity and aboutness (Sluckin et al. Reference Sluckin, Cruschina, Martine, Meklenborg and Wolfe2021); yet, V3 is unconstrained by argument structure, as transitive verbs are common (7c,d). Kiezdeutsch V3 instead involves a broader requirement in the preverbal position that permits subject DPs and spatio-temporal adverbials, while potentially excluding accusative object DPs or other unattested adverb(ia)ls, such as manner adverbs. Thus, a simple EPP-style subject requirement is inappropriate. We now turn to two topic-based approaches before testing their predictions.

4.2 A familiar topic requirement in V3

Freywald et al. (Reference Freywald, Leonie Cornips, Nistov, Opsahl, Nortier and Svendsen2015) and Walkden (Reference Walkden2017) consider the preverbal XP to be a familiar topic, a crucial ingredient in Walkden’s formal analysis (see section 6). A familiar topic is a D(iscourse)-linked, “given or accessible [i.e. discourse salient] (cf. Chafe Reference Chafe and Li1976) constituent, which is typically de-stressed and realized in a pronominal form” (Pesetsky Reference Pesetsky, Reuland and ter Meulen1987), often used for providing topic continuity (Givón Reference Givón1983) (see also Frascarelli & Hinterhölzl Reference Frascarelli, Hinterhölzl, Schwabe and Winkler2007:88), yet when they are not D-linked they must refer to a non-new aboutness topic (Schwabe & Winkler Reference Schwabe and Winkler2007:22).

Frequent pronominal subjects in Kiezdeutsch V3 chime with a familiar topic, as pronouns are often inherently D-linked, that is they resume antecedents in the discourse. However, this hypothesis predicts, if correct, that newly introduced indefinite subjects should be incompatible with V3. Furthermore, V3 should be ruled out in out-of-the-blue contexts. We test these predictions in corpus data in section 4.4 so as to assess their appropriateness. We now consider the position that V3 is related to preverbal sentence topics (Wiese Reference Wiese2006, Reference Wiese2009, Reference Wiese, Abraham and Liess2013).

4.3 A sentence topic requirement

Wiese (Reference Wiese2006, Reference Wiese2009, Reference Wiese, Abraham and Liess2013) claims that the preverbal position is instead conditioned by a broad information- structural requirement for topics (see also Wiese et al. Reference Wiese, Freywald and Mayr2009; Schalowski Reference Schalowski2015, Reference Schalowski2017; Wiese & Rehbein Reference Wiese and Rehbein2016; Wiese & Müller Reference Wiese, Müller, Antomo and Müller2018), which she defines as a Sentence Topic. However, Wiese and colleagues do not posit a formalized restriction on argument types (beyond noting the unavailability of preverbal objects), nor have they articulated more fine-grained restrictions. We briefly discuss the notion of a sentence topic before considering associated predictions. We ultimately argue that this characterization is too broad from a syntactic perspective, demonstrating the challenges of characterizing the preverbal element in Kiezdeutsch V3.

Sentence topichood corresponds to the idea of a topic-comment distinction (Reinhart Reference Reinhart1981, Gundel Reference Gundel1985), in which the topic is what the sentence is about (Reinhart Reference Reinhart1981, Lambrecht Reference Lambrecht1994); note the definition as given by Gundel (Reference Gundel1985):

Thus, a sentence topic is neither obligatorily given nor retrievable from the context, although it can be so additionally. It can be “newly introduced, newly changed or newly returned to” (cf. Givón Reference Givón1983:8), relating simply to the notion of aboutness, which is more basic than D-linking and givenness (see Rizzi Reference Rizzi, Laura Brugè, Munaro, Schweikert and Turano2005, Reference Rizzi, Petrosino, Cerrone and van der Hulst2018); a notational variant of sentence topic is thus aboutness topic. Indeed, aboutness/sentence topics can be indefinite, which are by definition discourse-new, as claimed by (Frey Reference Frey2004a) for a sentence like (9).

A preference for preverbal subjects, but not restriction, is unsurprising if there is a requirement for preverbal sentence topics in V3. However, how does this chime with the observation that the preverbal topic in Kiezdeutsch V3 cannot be an accusative object DP? Several contexts in SG permit preposed accusative objects as sentence/aboutness topics; these include generic utterances such as (10a) or left dislocated DPs (10b).

Therefore, a sentence topic requirement for preverbal XPs in Kiezdeutsch V3 does not fully encompass the preverbal element. However, this is an empirical issue that can be checked in corpus data.

4.4 New data for the preverbal position

We searched for V3 in the main multiethnic KiDKo corpus (KiDKo-mu) (Rehbein et al. Reference Rehbein, Schalowski and Wiese2014) (c.345,000 tokens and 23,506 matrix clauses), using the syntactic query used by Walkden (Reference Walkden2017) and a PoS-based query from the corpus handbook (Bunk & Blevins Reference Bunk and Blevins2017) (see the Appendix for all queries).

We found 199 instances of V3, more than previous studies. Overall, the reported tendencies held up; most preverbal XPs are subjects and most are D-linked pronouns. However, we found multiple instances of preverbal subjects that cannot be familiar topics, for example 4 cases of impersonal pronouns which are by definition not D-linked (11a), 1 nonreferential pseudo-argument es (11b), and 8 indefinite subjects (11c).

In sum, these data challenge a familiar topic requirement (contra Freywald et al. Reference Freywald, Leonie Cornips, Nistov, Opsahl, Nortier and Svendsen2015, Walkden Reference Walkden2017). From a purely informational perspective, Wiese’s sentence topic generalization (Wiese Reference Wiese2006, Reference Wiese2009, Reference Wiese, Abraham and Liess2013) is more appropriate. Yet, the restriction against preverbal accusative object DPs largely holds up, which cannot be explained by a sentence topic requirement. We note, however, that one speaker (Mu9WT) is much more liberal with the preverbal position in V3, accounting for very few counterexamples, e.g. a reported V4 preverbal light object pronoun (12a) as also reported by Schalowski (Reference Schalowski2017) and Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023) and an object-WH-pronoun (12b). We follow Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023) in assuming the possibility of some level of microvariation.

4.5 The preverbal XP as Subject of Predication

We argue that a more appropriate characterization of the preverbal position in Kiezdeutsch V3 relates to the notion of Subject of Predication (SoP) (see Cardinaletti Reference Cardinaletti and Rizzi2004, Rizzi Reference Rizzi, Laura Brugè, Munaro, Schweikert and Turano2005, Reference Rizzi, Petrosino, Cerrone and van der Hulst2018, Bentley & Cruschina Reference Bentley and Cruschina2018, Sluckin et al. Reference Sluckin, Cruschina, Martine, Meklenborg and Wolfe2021). Cardinaletti (Reference Cardinaletti and Rizzi2004) proposed that [±SoP]-bearing XPs can move to a position straddling the TP and CP domains (Subj(ect)P), which shows flexible properties akin to but not equal to both a subject and a topic position (see also Alexiadou & Anagnostopoulou Reference Alexiadou and Anagnostopoulou1998, Miyagawa Reference Miyagawa2017). However, Cardinaletti (Reference Cardinaletti and Rizzi2004) shows that the types of argument encoding SoP are limited; specifically, accusative DPs cannot be SoP. Moreover, from a pragmatic perspective, Rizzi (Reference Rizzi, Laura Brugè, Munaro, Schweikert and Turano2005:212) has argued that SoP equates to a sentence/aboutness topic, formalizing SoP as simply [+Aboutness], while given topics encode [+Aboutness, +D-linking]. We first critically introduce SoP, before applying the observations to Kiezdeutsch.

An important question then relates to what can and cannot be SoP. Cardinaletti (Reference Cardinaletti and Rizzi2004) examines Italian and wider Romance, Germanic, and Semitic data, showing that beyond nominative DPs and (noncliticized) pronouns, certain nonnominative XPs can occupy a subject position, which she calls SubjP between T and C. Alternatively, SubjP can be viewed as crosslinguistically variable properties of T or C, along with other discourse features, depending on the language at hand (Mohr Reference Mohr2005, Miyagawa Reference Miyagawa2017, Sluckin Reference Sluckin2021), a position we ultimately share. A detailed critique of SubjP is beyond our goals (but see Sluckin Reference Sluckin2021:§4). The following elements can act as SoP in Italian and beyond (where possible, we use aux-to-comp raising, demonstrating a TP-domain position; cf. Cardinaletti Reference Cardinaletti and Rizzi2004): nominative subject DPs and pronouns (13a), spatio-deictic adverbials/complements (13b), and (free) dative experiencers of unaccusatives and type III psych verbs, such as piacere ‘to please’; see Belletti & Rizzi Reference Belletti and Rizzi1988 (13c,d). A nonselected dative PP of an unergative is also marginally possible with a late focused subject DP (Cardinaletti Reference Cardinaletti and Rizzi2004:124) (13e). Notably, accusative object experiencers of Type II psych-verbs, such as preoccupare ‘to worry’ type, cannot be SoP. Recent work has argued that SoP is encodable on covert locatives (shown with an unergative in (13f)) and both covert and overt nonthematic situational arguments or situation-anchoring temporal adverbs (13g,h), such as German da ‘there’ (Bentley & Cruschina Reference Bentley and Cruschina2018, Sluckin et al. Reference Sluckin, Cruschina, Martine, Meklenborg and Wolfe2021, Cognola Reference Cognola2023).

Internal arguments of transitive verbs, that is accusative object DPs and also dative object DPs of ditransitives, cannot be SoP (14a,b).

Overall, Italian has a more flexible subject position than languages with strict nominative DP-subject requirements, such as English. This position possesses properties of both A and A’ positions. However, the ungrammaticality of (14a,b) shows that argument structure or Case properties influence what can be SoP. We now discuss the notion of SoP in more detail before viewing Kiezdeutsch V3 from this perspective.

4.6 Dissecting the SoP requirement

We briefly explore the nature of SoP and an SoP requirement, focusing chiefly on the Italian data underlying Cardinaletti’s (2004) observations. Rizzi’s (2005, 2018) suggestion that SoP is an information-structural notion related purely to aboutness cannot alone explain why internal DP arguments of mono- and ditransitives cannot be SoP; recall that accusative DPs can be aboutness topics in German. However, Cardinaletti’s (2004) suggested formal [±SoP] endowment requires dubiously defined semantic limitations on syntactic formal features assigned to arguments.Footnote ²

The most pertinent observation is that accusative DPs cannot be SoPs, indicating a relationship between structural Case and the subject position. Indeed, Cardinaletti (Reference Cardinaletti and Rizzi2004) considers limitations on SoP to be understandable in terms of a difference between inherent and structural Case, namely object DPs with structural accusative Case cannot be SoP but inherent datives can; yet for Cardinaletti, SubjP is not a Case-assigning position. Since accusative object experiencers are assigned inherent Case (Belletti & Rizzi Reference Belletti and Rizzi1988, Landau Reference Landau2010), they should then in theory be possible, but they are not. If accusatives are universally prohibited but datives are not, the subject position must relate to nominative Case assignment (contra Cardinaletti Reference Cardinaletti and Rizzi2004:126), akin to T assigning nominative Case (Chomsky Reference Chomsky1995). Crucially, structural Case positions should be blind to inherent Case, which is assigned via separate mechanisms before PF (McFadden Reference McFadden2004, Woolford Reference Woolford2006, Landau Reference Landau2010). So why are inherent accusative DPs impossible as SoP? We suggest that all accusative DPs are incompatible with nominative-assigning subject positions due to interpretive ambiguity at the syntax–phonology interface, which cannot differentiate the matching exponence of structural and inherent accusative Case. Thus, nominative-assigning properties clash with accusative-marked XPs and prevent Convergence at the interfaces, leading to Crash (Chomsky Reference Chomsky1995). We expand the specific mechanism when discussing Kiezdeutsch in section 7.

The disparity regarding different dative arguments is more puzzling. Why can dative experiencers be SoP but not the internal dative DP arguments of ditransitives? Although Italian dative experiencers are considered quirky subjects (see Belletti & Rizzi Reference Belletti and Rizzi1988, Belletti Reference Belletti, Mirko Grimaldi, Franco and Baldi2018), unlike dative indirect objects, both receive inherent Case assigned to a thematic position by a v-head (Belletti & Rizzi Reference Belletti and Rizzi1988, McFadden Reference McFadden2004, Landau Reference Landau2010). Notably, class III psych-verbs are a subtype of unaccusative in which the experiencer is the highest DP and closest to the subject position. If the varying behavior of dative arguments cannot be explained via Case, the dative internal arguments of ditransitives must instead be blocked by the nominative external argument; Cardinaletti (Reference Cardinaletti and Rizzi2004:124) notes that PP scrambling over an external subject argument is never possible. Thus, when there is a choice between a vP external and an internal DP argument, the former is always a more appropriate goal.

The described locality effects appear related to the unavailability of scrambling in VO languages (Haider Reference Haider2010): dative indirect objects have no space to leapfrog an external argument within the TP-domain. However, some reordering in the vP-edge is possible (Belletti Reference Belletti and Rizzi2004); yet, solving Italian low-reordering is beyond our goals. Nonetheless, SoP evades locality requirements in a few contexts. For example, the lower nominative theme of a psych-verb can topicalize to the subject position unproblematically when the predicate remains low under narrow or broad focus. Moreover, in unergative contexts, fronted nonselected (experiential) dative PPs and covert locative/deicitc arguments (see Pinto Reference Pinto1997, Sluckin et al. Reference Sluckin, Cruschina, Martine, Meklenborg and Wolfe2021) can raise above the vP external argument to the subject position; yet again these instances relate to broad focus on the vP complex or narrow focus on the external argument so that the nominative DP subject remains low (see Pinto Reference Pinto1997, Bentley & Cruschina Reference Bentley and Cruschina2018, Sluckin et al. Reference Sluckin, Cruschina, Martine, Meklenborg and Wolfe2021).

Overall, the so-called SoP position is visibly sensitive to Case and locality, beyond aboutness (Rizzi Reference Rizzi, Laura Brugè, Munaro, Schweikert and Turano2005, Reference Rizzi, Petrosino, Cerrone and van der Hulst2018), regardless of approaches assuming a Cartographic array (Cardinaletti Reference Cardinaletti and Rizzi2004), variations on minimalist Spec,TP (e.g. Chomsky Reference Chomsky1995, Miyagawa Reference Miyagawa2017), or split-INFL (Pollock Reference Pollock1989, Alexiadou & Anagnostopoulou Reference Alexiadou and Anagnostopoulou1998). The SoP position is neither an unrestricted topic position nor a strictly D-related subject position, showing both A and A’ properties. Aside from the validity of SubjP, greater flexibility in Spec,TP in null-subject languages is predicted via different mechanisms, such as D-satisfaction via V-to-T movement (Barbosa Reference Barbosa1995, Alexiadou & Anagnostopoulou Reference Alexiadou and Anagnostopoulou1998) and/or an SoP requirement on T itself (Mohr Reference Mohr2005, Sluckin Reference Sluckin2021). We thus generalize that this position is related not only to aboutness but also to nominative Case assignment, a property incompatible with accusative DPs; but SoPs can be nominative, dative (with some restrictions), or C/caseless, such as a locative or situational adjunct/argument. This behavior is expected where D properties of T are satisfied via mechanisms alternative to DP-movement. Hence, we conclude that the syntactic notion of SoP is epiphenomenal. That is, if information structure, C/case, and locality determine in tandem what can and cannot be SoP, then SoP is a portmanteau property. Overall, we sum up the generalizations as follows:

1. i. A syntactic SoP may bear structural nominative Case, inherent dative Case, or no Case.

2. ii. A syntactic SoP cannot bear structural or inherent Accusative; the latter is ambiguous with the former at PF.

Let us now discuss how an SoP-position is descriptively appropriate for Kiezdeutsch V3 without yet providing a formal analysis.

4.7 Preverbal elements in Kiezdeutsch V3 as SoP

The above description of SoP encompasses most of the reported preverbal elements in Kiezdeutsch V3, namely nominative subject DPs and pronouns, and some temporal and locative adverbs and arguments (e.g. da, dann as shown above), which can all be understood as SoP. While dative experiencers in V3 should then be possible, we find only one example of a free dative (15), albeit with a correction limiting its meaningfulness. Given the limited corpus size and the problem of negative evidence in such research, a full realization of predicted categories could remain elusive.

We note that we find no instances of fronted accusative object experiencers of the type in (16) in V3, aligning with predictions based on Italian and what we know about Kiezdeutsch V3. Indeed, beyond V3, we find no evidence at all for such inherent accusative object experiencers at all. Although, one might consider them SoPs (Mohr Reference Mohr2005) at least in the Aristotelian sense. We return to a short discussion of inherent accusative object experiencers in Kiezdeutsch and SG in section 7.

A reviewer asks if nonreferential quasi-argument es in existential constructions (11b) can be SoP. We draw on work by Hinterhölzl (Reference Hinterhölzl2019) and Fuß & Hinterhölzl (Reference Fuß and Hinterhölzl2023) on expletive elements, who view quasi-argumental es as a weak demonstrative item which is a generalized quantifier corresponding to a situational argument, that is, es denotes a property of the situation and can be understood as a situation topic which anchors the discourse in a reference situation. Thus, for Fuß & Hinterhölzl (Reference Fuß and Hinterhölzl2023), es has an existential impact on the assertion of the speaker and naturally anchors an existential statement in the context. Following the aforementioned literature arguing that situational arguments may encode SoP (Bentley & Cruschina Reference Bentley and Cruschina2018, Sluckin et al. Reference Sluckin, Cruschina, Martine, Meklenborg and Wolfe2021, Sluckin Reference Sluckin2021, Cognola Reference Cognola2023), we hold that pseudo-argumental es is an SoP (11b). In short, es encodes aboutness concerning the situation but is not D-linked.

In sum, the incompatibility of the preverbal position in V3 and accusative DPs indicates an active SoP requirement in the Kiezdeutsch prefield. We do not yet formalize the exact syntactic nature of Kiezdeutsch’s SoP, yet in section 7 we will argue that the necessary features are bundled in a low position in the C-domain. Furthermore, we must also ask how Kiezdeutsch produces V2 O-V_FIN-S matrix clauses if the preverbal position is reserved for SoPs. We assume that non-SoP XPs target a higher position in the C-domain (see also Walkden Reference Walkden2017) (sections 6 and 7). Given the strictness of SG V2 and the well-known lack of preferential treatment of nominative subjects in the left periphery (den Besten Reference den Besten and Abraham1983, Vikner Reference Vikner1995), we posit that it lacks the necessary ingredients for an SoP requirement (den Besten Reference den Besten and Abraham1983, Vikner Reference Vikner1995). Finally, if an SoP-requirement holds in Kiezdeutsch, it should manifest beyond Adv-XP-V_FIN V3, for instance, as a lack of preverbal object DPs in resumptive Left Dislocation and Hanging Topic Left Dislocation. This is an empirical matter to which we now turn.

5.1 Differentiating types of resumption

LD typically involves adjacent resumption of an integrated dislocated DP argument by a resumptive d-pronoun in the preverbal position, agreeing in ϕ-features and Case with the dislocate (18a, b), or resumption of an adverbial by a resumptive adverbial pronoun, such as da, dann, danach, so (18c) (Altmann Reference Altmann1981, König & Auwera Reference König, van der Auwera, Haiman and Thompson1988, Grohmann Reference Grohmann1997, Reference Grohmann, Billerey and Danielle Lillehaugen2000, Reference Grohmann2003, Zifonun et al. Reference Zifonun, Hoffmann, Strecker and Ballweg1997, Frey Reference Frey, Lohnstein and Trissler2004b, Shaer & Frey Reference Shaer and Frey2004).

HTLD involves resumption of an unintegrated adjunctival dislocate (Grohmann Reference Grohmann2003) via a p-pronoun in the preverbal position (19a) or can be discontinuous with either a p- or d-pronoun in the middlefield (19b). We understand adverbial resumption in the middlefield as HTLD (18c). Requirements for Case and ϕ-agreement are weakened (Altmann Reference Altmann1981, Grohmann Reference Grohmann2003, Frey Reference Frey, Lohnstein and Trissler2004b, Shaer & Frey Reference Shaer and Frey2004, Petrova Reference Petrova2012) leading to some optionality in Case agreement between the hanging topic (HT) and the resumptive pronoun (henceforth RP) (see also Samo Reference Samo2019), as in (19b).

Although HTLD and LD can be ambiguous (Altmann Reference Altmann1981, Grohmann Reference Grohmann2003), HTLD often involves a prosodic break between the dislocate and RP, while LD dislocates and RPs are prosodically inseparable. However, prosody can be unreliable (Frey Reference Frey, Lohnstein and Trissler2004b) and cannot be tested effectively in corpus data. We thus maintain the definitions above.

5.2 A quantitative corpus study of fronted XPs in dislocation

We investigated the distribution of subject and object DPs in resumption in Kiezdeutsch. We searched KiDKo-mu (Rehbein et al. Reference Rehbein, Schalowski and Wiese2014) for both HTLD and LD involving DPs, and adverbials for completeness. We employed queries with both syntactic annotation and Parts-of-Speech (POS) tagging in order to catch potentially erroneously tagged tokens (see Appendix for queries). We manually checked all results to distinguish arguments and excluded false hits and divided them according to metadata on home language.

Adverbials outnumbered DPs in LD (table 1) (see also Sluckin & Bunk Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023). Notably, preverbal objects are rare, 2.3 percent (n=6) of all LD. Moreover, when we consider the metadata on home language, 5 hits are from speakers from German-speaking homes; one other comes from an interlocutor for whom we lack metadata.

Table 1. Types of LD in Kiezdeutsch across speaker types

HTLD was less frequent than LD but also shows a strong preference for preverbal subjects (91 percent) regardless of the dislocate XP (table 2).

Table 2. The distribution of possible HTLD orders in Kiezdeutsch

We find only one preverbal dative argument across LD and HTLD (20), an experiencer subject, produced by a multilingual speaker.

When results are broken down according to home language (see table 3), all O>VFIN comes from speakers from German-speaking homes.

Table 3. The distribution of preverbal DP types in HTLD across Kiezdeutsch speaker groups

Strikingly, the combined results for LT/HTLD in table 4 show that 9/10 preverbal objects come from speakers from German-speaking homes.

Table 4. The distribution of preverbal elements in all resumption in Kiezdeutsch across speaker groups

Given the limited data, statistical analysis was conducted on the combined LD/HTLD data comparing fronted objects and subjects for the speaker subgroups according to the metadata on home language (or lack thereof for interlocutors).Footnote ³ The results are given in table 5 with significant results in boldface.

Table 5. Statistical comparison of preverbal objects and subjects in resumption according to metadata for home language in KiDKo-mu

The tests in table 5 reveal a highly significant difference between the two anchor groups’ usage of preverbal object and subject RPs, that is, speakers with and without German as a home language appear to behave differently. Furthermore, interlocutors also differ significantly from the monolingual anchor speakers but not from the multi/bilinguals in this respect. Thus, the results indicate, albeit tentatively, that interlocutors and speakers from multilingual backgrounds avoid preverbal objects more than their peers from monolingual households.Footnote ⁴

In sum, the differences between speaker groups pertain to frequency and, by statistical inference, availability of fronted objects. Thus, the behavior of argument fronting in Kiezdeutsch resumption strategies shares parallels with the V3 pattern, most markedly in the speech of speakers who are unlikely to speak German at home.

5.3 Comparison with monolingual spoken German in KiDKo and Tüba-D/S

For comparison, we examine resumption beyond Kiezdeutsch, investigating the relative distributions of argument types from adolescent monolingual/ethnic speakers in the smaller KiDKo-mo companion corpus (Rehbein et al. Reference Rehbein, Schalowski and Wiese2014) (147,000 tokens, 8945 matrix clauses) and adult monolingual speakers in the Tübinger Baumbank des Deutschen/Spontansprache (TüBa-D/S) ‘Tübinger tree bank of German/spont-aneous speech’ (Hinrichs et al. Reference Hinrichs, Julia Bartels, Kordoni, Telljohann and Wahlster2000) (c. 360,000 tokens, 28,545 matrix clauses). Significance tests are applied to these data sets below

In KiDKo-mo, we find a strong preference for preverbal subjects in LD and HTLD of DP arguments, similar to Kiezdeutsch. However, the sample size was considerably smaller than the two other comparably sized corpora KiDKo-mu and TüBa-D/S. Overall, we found 29 examples LD involving DPs, one more than Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023), (table 6)Footnote ⁵ and 8 such instances of HTLD (table 7). Only 3 hits of LD and 1 of HTLD showed preverbal object DPs.

Table 6. Types of LD in monoethnic youths in KiDKo-mo

Table 7. Types of HTLD in monoethnic youths in KiDKo-mo

The combined results (table 8) show a similar overall distribution to KiDK-mu. Yet, while these speakers produced more fronted objects than the confirmed multilingual speakers and interlocutors in KiDKo-mu, the limited data and smaller corpus limit the meaningfulness any differences. Notably, speakers from German-speaking homes in KiDKo-mu produced more preverbal objects than speakers in KiDKo-mo; however, some data is too sparse for tests of statistical significance.

Table 8. The distribution of preverbal elements in all resumption in monoethnic youths in KiDKo-mo

Compared to the KiDKo corpora, the more standard spoken data from TüBa-D/S showed more frequent preverbal objects in LD (c.14.1%) and HTLD (13%); We used the syntactic annotation-based query [cat=“LV”], as HTs and LD topics are not annotated differently. Nonetheless, preverbal subjects were the most frequent preverbal D-element (40.6% in LD and 70.1% in HTLD), suggesting a general S>V_FIN tendency in DP resumption; this is unsurprising as S>V_FIN is more common than O>V_FIN in German matrix clauses (50% S>V_FIN, 7% O>V_FIN, 42% Adv>V_FIN, 1% other; cf. Bohnacker & Rosén Reference Bohnacker and Rosén2008:517). Notably, fronted objects were more common in dislocation in TüBa-D/S than is reported for declarative matrix (see Bohnacker & Rosén Reference Bohnacker and Rosén2008). Given that TüBa-D/S contains many question-answer-based interactions, we find some HTLD with preverbal C-elements, for example, wh-words such as wie ‘how’; this is absent in the KiDKo corpora. Overall, we find greater variation in HTLD types in TüBa-D/S. The distinct types of resumption are broken down according to dislocate and preverbal XP for HTLD in table 9 and for LD in table 10. The combined relative distributions of fronted XPs in LD and HTLD are summarized in table 11.

Table 9. Frequency of attested HTLD types in TüBa-D/S

Table 10. LD in Tüba-D/S

Table 11. Frequency of preverbal XPs in re sumption from TüBa-D/S

Turning to statistical analysis of these data, comparison of fronted objects and subjects in resumption across KiDKo-mu, KiDKo-mo, and TüBa-D/S is somewhat inconclusive (see table 12). Notably, however, there is a significant difference between the KiDKo-mu and TüBa-D/S, namely, Kiezdeutsch and standard-aligned adult spoken German. In contrast, differences between the two KiDKo corpora and between KiDKo-mo and TüBa-D/S are not significant; the considerably smaller size of KiDKo-mo inhibits effective testing. Thus unsurprisingly, Kiezdeutsch speakers diverge most from spoken adult SG.

Table 12. Statistical comparison of corpus populations regarding differences in the relative distributions of preverbal subject and object DPs in resumption

We then compared the individual subgroups in KiDKo-mu according to home language (Ger(man) or Her(itage)) or inter(locutor) status with KiDKo-mo and TüBa-D/S. The results are given in table 13 with significant results in boldface.

Table 13. Statistical comparison of KiDKo-mu samples according to home language with KiDKo-mo and TüBa-D/S corpus populations regarding the relative distributions of preverbal subject and object DPs in resumption

In sum, comparison of KiDKo-mu and TüBa-D/S delivers some striking results: (i) speakers from German-speaking homes do not differ significantly from adult speakers; (ii) the differences between TüBa-D/S and the KiDKo-mu samples from the interlocutors and anchors from non-German-speaking homes are highly significant (p < 0.01). However, no significant differences were present between the Kiezdeutsch subgroups and the monoethnic companion corpus. Thus, it seems that younger speakers employ fewer preverbal objects in resumption, but this trend is strongest in Kiezdeutsch speakers from non-German-speaking backgrounds. This strengthens the finding that such Kiezdeutsch speakers behave differently from those from German-speaking homes.

5.4 Discussion

The results support observations that adult speakers of SG, monoethnic adolescents, and Kiezdeutsch-speaking multiethnic adolescents employ the same categorial range of preverbal elements in resumption (Sluckin & Bunk Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023), namely, DPs and adverb(ial)s. The lack of wh-words in HTLD in KiDKo-mu/mo may result from the different types of interaction. However, Kiezdeutsch speakers from non-German-speaking homes and the interlocutors appear to avoid resumptive preverbal objects, which is uncharacteristic of both monoethnic adults or Kiezdeutsch speakers from German-speaking homes. As noted, the KiDKo-mo data were inconclusive. Most notably, 90 percent of preverbal object RPs in the Kiezdeutsch data come from 3/4 anchor speakers from German-speaking homes. This finding for speakers from multilingual/ethnic backgrounds can be taken as evidence for a more pervasive SoP requirement in the prefield of their grammars.

However, some factors mitigate the results: firstly, we lack metadata for interlocutors’ home languages, although based on Wiese et al.’s (Reference Wiese, Freywald, Schalowski and Mayr2012) demographic data, many will come from non-German-speaking homes; the performance data supports this position. Secondly, negative evidence is unavailable in corpora, yet an acceptability study is problematic, as (i) the anonymous speakers (now aged 28–31) cannot be followed up; (ii) the current adolescent population, the next generation, is not a like-for-like comparison; and (iii) blurred boundaries between varieties and speakers’ proficiency in Kiezdeutsch and SG means that SG grammar is likely always acceptable. Nonetheless, some microvariation in Kiezdeutsch is apparent (see also Sluckin & Bunk Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023), which appears conditioned by linguistic background. We speculate that this relates to differences in the timing and level of exposure to German in and outside the home (see also Walkden Reference Walkden2017), and variation in the linguistic behavior of peers in early childhood; We address the nature of bilingual CLA and a potential innovation scenario in section 9. Kiezdeutsch speakers from German-speaking homes’ different behavior results then from either microvariational parametric differences or an e-language phenomenon, whereby they adopt V3 as a sociolinguistically salient pattern; indeed, Wiese (Reference Wiese2009:790) reports a school teacher’s impression that use of salient Kiezdeutsch forms increases around age 12; yet, it is unlikely that speakers of Kiezdeutsch would seize on a pattern as a stereotyped in-group marker from more general SG-aligning spoken language. If speakers only adopt V3 during adolescence, V3 is reduced to an e-language phenomenon, which is not the general consensus. We refrain from stronger conclusions for these speakers.

The absence of fronted accusative direct objects DPs in the V3 and resumption of multilingual Kiezdeutsch speakers is conspicuous such that we consider a wider syntactic constraint desirable. If Kiezdeutsch differed from SG due to a structural requirement for familiar topics in the preverbal position, as argued by Walkden (Reference Walkden2017), more preverbal object resumptive pronouns are expected, since RPs are familiar topics par excellence. It is unlikely, however, that the marked lack of preverbal accusative objects in both V3 and resumption falls out from a tendency for subjects to more often be topics than objects. Instead, this result, namely, a pervasive aversion to preverbal accusative object DPs across Kiezdeutsch verb-third structures, appears to follow from the predictions made in section 4 if the preverbal position across all V3 in Kiezdeutsch is indeed associated with SoP properties. However, we also found hardly any dative experiencers in the preverbal position. Thus, the results for multilingual speakers correspond to a subset of possible SoPs, although it is unclear if a rule preventing preverbal dative XPs in V3 is desirable; yet both instances were produced by multilinguals. We now turn to a syntactic analysis for Kiezdeutsch which attempts to capture the empirical findings.

6.1 Cartographic approaches

In recent years, cartographic approaches using a Split-CPs (see Rizzi Reference Rizzi and Haegeman1997) have been em- ployed to model V2 and V2> word orders in historical and contemporary V2 varieties across Germanic and Romance (see, for example, Poletto Reference Poletto, Barbiers, Cornips and van der Kleij2002, Walkden Reference Walkden2014, Wolfe Reference Wolfe2015, Reference Wolfe2018, Reference Wolfe2019, Hinterhölzl Reference Hinterhölzl2017, Haegeman & Greco Reference Haegeman and Greco2018, Samo Reference Samo2019, Greco & Haegeman Reference Greco, Haegeman, Woods and Wolfe2020, Meelen et al. Reference Meelen, Mourigh, Lai-Shen Cheng, András Bárány, Douglas and Vikner2020, Breitbarth Reference Breitbarth2022, Reference Breitbarth2023, Sluckin & Bunk Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023). Much work has focused on the information structural properties of XPs in the prefield and the position targeted by the verb as either a low or high head in Split-CP, namely, Fin or Force. A high position in Force should in theory leave fewer available positions above the Vfin leading to a stricter V2 system, while a lower locus of V_FIN should enable a more flexible system with more possible V3> orders (21).

For due diligence, we introduce a brief overview of these approaches, before setting out why we do not adopt them. For Kiezdeutsch specifically, two cartographic approaches have been suggested; Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023) follow a Wolfian approach (Wolfe Reference Wolfe2015, Reference Wolfe2018, Reference Wolfe2019), proposing that both German and Kiezdeutsch are strict V2 systems with a high locus of VFIN with a high Frame-field (Benincà & Poletto Reference Benincà, Poletto and Rizzi2004) above ForceP which may host HTs, LDed dislocates, and framesetters. The key difference between Kiezdeutsch and SG for them is then the locus of framesetting adverbials in each of these varieties. Specifically, they propose that SG merges these elements below Force, while Kiezdeutsch patterns with a range of V3 varieties that have been proposed to merge them in FrameP which sits above ForceP (see approaches specifically by Haegeman & Greco Reference Haegeman and Greco2018, Wolfe Reference Wolfe2019, Greco & Haegeman Reference Greco, Haegeman, Woods and Wolfe2020, Meelen et al. Reference Meelen, Mourigh, Lai-Shen Cheng, András Bárány, Douglas and Vikner2020); see below:

Indeed, recent work by Breitbarth (Reference Breitbarth2023) finds very marginal, yet systematic and innovative availability of high central adverbs, for example, jetzt ‘now’ in V3 in L1 colloquial German. These apparently Merge in a high position akin to Hanging Topics and thus also framesetters. This phenomenon has been analyzed similarly for Kiezdeutsch by Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023). Given the functional overlap between framesetters and temporal central adverbials, we ask if the innovation may not only come down to Merge site but also to the role of the adverbs, i.e. it is possible that such innovation equates to a functional merger.

Alternatively, Hinterhölzl (Reference Hinterhölzl2017) has suggested a typological difference between V2 grammars with V3, such as Kiezdeutsch (which he mentions only passing), and those without, such as SG. Hinterhölzl’s argument is based on a specific approach he develops for phase theory, including the determination of the phase head, the rigidity/flexibility of the phase edge, and the phase’s integration with prosody, and the mapping of different element types into particular phonological phrases. Firstly, Hinterhölzl (Reference Hinterhölzl2017) assumes a different conception of Split-CP without a Force-external frame-field (cf. Frascarelli & Hinterhölzl Reference Frascarelli, Hinterhölzl, Schwabe and Winkler2007, Speyer Reference Speyer2008) (23); thus, framesetters (or F(rame)-Topics in his terms) occupy the position immediately below ForceP.

Secondly, and in short, SG is subject to the prosodic condition in (25); thus, such that the phase edge in a strict V2 language like SG is located on ForceP; this in turn forces movement of V_FIN as high as Force and indeed some other XP to Spec,ForceP, for instance, the F-Topic directly below Force. However, if we have understood correctly, such a prosodic condition is not active in Kiezdeutsch and thus V_FIN moves to Fin but there is no requirement that V_FIN move to Force. This means that Kiezdeutsch has only a low bottleneck on FinP, but elements generated in higher projections can lead V3> orders.

Therefore the formal difference between SG and Kiezdeutsch can be understood in terms of the structures in (25a,b).

Hinterhölzl’s model can then quite easily explain V3 in Kiezdeutsch. Likewise, the model appears compatible with LD in Kiezdeutsch under different movement analyses (Grohmann Reference Grohmann1997, Grewendorf Reference Grewendorf2002b,a, Reference Grewendorf, Benjamin Shaer, Frey and Maienborn2009) assuming movement of a DP-structure through Spec,FinP. On the other hand, a prosodic condition forcing Fin-to-Force movement seemingly might be problematic for dislocation structures in SG; while we admit that Hinterhölzl (Reference Hinterhölzl2017) does not directly address LD and HTLD, we wonder how it can deal with LD specifically. That is, assuming that LD is phonologically integrated and V_FIN sits on Force, one might assume that a phonologically integrated big-DP is moved in its entirety to Spec,ForceP, rather than the full DP undergoing extraction to a higher phrase. This raises the question as to why resumption is necessary at all, if both XPs remain in the same projection; we return to this problem in our own analysis in section 7, arguing that a doubled argument in an unsplit big-DP would be deleted at PF. Moreover, it seems that HT can only occur as a syntactically unintegrated clausal adjunct for Hinterhölzl (Reference Hinterhölzl2017) in both SG and Kiezdeutsch; however, Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023) report limited instances of framesetters preceding diagnosable HTs. This is problematic for Hinterhölzl’s model, although his theoretically ambitious paper is not large enough to discuss every possible eventuality; these issues may have easy fixes. However, if Hinterhölzl (Reference Hinterhölzl2017) is correct, Breitbarth’s (Reference Breitbarth2023) findings suggest that the prosodic condition in (24) could be weakening. However, we do not ultimately adopt Hinterhölzl’s or any cartographic system for a range of reasons. We adopt more minimalist assumptions with the exception that we permit information structural formal features (Miyagawa Reference Miyagawa2017).

From a technical perspective, we assume the minimalist position that phase heads come syntactically predetermined (Chomsky Reference Chomsky and Kenstowicz2001, Reference Chomsky, Freidin, Otero and Luisa Zubizarreta2008, Ouali Reference Ouali, Roberta D’Alessandro and Hrafn Hrafnbjargarson2008, Reference Ouali2011, Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010) and are responsible for feature inheritance to a lower functional head, namely, C to T. In the first instance, we are unsure how cartographic approaches roughly adopting the mantra of “one feature one head” (1F1H) are truly compatible with inheritance;Footnote ⁶ of course, cartographic conceptions of the phase and the strength of this mantra may vary. Indeed, Bacskai-Atkari (Reference Bacskai-Atkari2023:31) points out that information-structural heads between Force and Fin are apparently independent of selection restrictions in the accounts by Rizzi (Reference Rizzi and Haegeman1997, Reference Rizzi and Belletti2004), yet a split-CP appears to require a Top head to select a FocP, a Foc head to Select another TopP and that Top head to select Fin; in turn this removes any selectional relationship between the core projections of ForceP and FinP. Another theoretical concern we have then relates to the fact that 1F1H would prohibit any account of C-selection which assumes mechanisms related to formal features (see discussions in Svenonius Reference Svenonius1994, Adger Reference Adger2003).Footnote ⁷

Returning to the individual approaches discussed above (Hinterhölzl Reference Hinterhölzl2017, Sluckin & Bunk Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023), we are unsure as to what the exact phonological or syntactic prerequisite for being a phase head is under Hinterhölzl’s account. Aside from the stipulation that some languages do or do not have a prosodic requirement for V-movement to the phase edge, it is unclear why Force is a phase head or edge in one language but not another; Hinterhölzl (Reference Hinterhölzl2017) provides an ambitious revision of phase theory, but we are unsure how it aligns with wider mainstream approaches to phasehood. Another potential issue, however, is that the low Fin-V2 system he suggests for Kiezdeutsch could predict unreported types of V4 orders (for such analyses of flexible V2 in medieval Romance see Wolfe Reference Wolfe2018, Reference Wolfe2019); yet this may boil down to variation in what can and cannot be base generated in the Germanic left periphery. Finally, neither Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023) nor Hinterhölzl (Reference Hinterhölzl2017) can derive the general prohibition on fronted accusative DPs which we find across V3, LD, and HTLD in Kiezdeutsch, for example, they do not rule out Frame>O>V_FIN, which we believe to be a syntactic desideratum in the grammars of multilingual Kiezdeutsch speakers.Footnote ⁸

6.2 Walkden’s double CP approach

We take Walkden’s (2017) feature-scattering approach as a base (see also Giorgi & Pianesi Reference Giorgi and Pianesi1997, Hsu Reference Hsu2017), deriving V3 from a double CP-structure, which Walkden stresses is not CP recursion in the sense of Iatridou & Kroch (Reference Iatridou and Kroch1992) and Vikner (Reference Vikner1995). Here, V_FIN targets a lower C-head, C¹, and a familiar topic targets its specifier; this is a conflation of Rizzi’s (1997) FinP and Frascarelli & Hinterhölzl’s (2007) low left-peripheral topic position FamP. Other active information-structural categories, such as framesetters, contrast, focus, shift (aboutness) topics, target a higher specifier on C², a conflation of Force and other information-structural projections, producing the V3 structure in figure 1.

Figure 1. Walkden’s (2017) Kiezdeutsch C-domain.

Walkden’s analysis encounters the established problem that a familiar topic requirement is pragmatically too restrictive (§4), nor can it exclude preverbal accusative objects. Walkden (Reference Walkden2017:61) supposes that the strictness of SG V2 is derived when all “the different possible left-peripheral projections are … syncretized into a single multifunctional … CP.” Thus, Kiezdeutsch has innovated a low topic requirement and expanded its CP structure. Yet a single CP could not straightforwardly derive LD or HTLD in SG. While we agree that most features relevant for SG V2 are concentrated on C¹, the presence of a higher CP is a theoretical require- ment for any analysis of LD and HTLD (Grohmann Reference Grohmann1997, Reference Grohmann, Billerey and Danielle Lillehaugen2000, Reference Grohmann2003, Grewendorf Reference Grewendorf2002a,b, Reference Grewendorf, Benjamin Shaer, Frey and Maienborn2009, Frey Reference Frey, Lohnstein and Trissler2004b, Boeckx & Grohmann Reference Boeckx, Grohmann and Abraham2005, den Dikken & Surányi Reference Dikken and Surányi2017, Sluckin et al. Reference Sluckin, Cruschina, Martine, Meklenborg and Wolfe2021), suggesting that the difference between SG and Kiezdeutsch is not the number of C-heads but the distribution of features across these heads. We return to the syntax of HTLD/LD below. We now incorporate the new data into a modified version of Walkden’s analysis.

7.1 Clashing Case explains an epiphenomenal SoP restriction

Recall that SoP has the pragmatic property of encoding aboutness (without d-linking) (Rizzi Reference Rizzi, Laura Brugè, Munaro, Schweikert and Turano2005, Reference Rizzi, Petrosino, Cerrone and van der Hulst2018, Bentley & Cruschina Reference Bentley and Cruschina2018, Sluckin et al. Reference Sluckin, Cruschina, Martine, Meklenborg and Wolfe2021) but shows a syntactic incompatibility with accusative Case. Based on Italian data, in section 4.5, we suggested that this (in)compatibility relates to nominative Case assigning properties of the SoP position, or rather TP (although not incompatible with Cardinaletti’s 2004 SubjP). We translate this into Kiezdeutsch by placing the SoP requirement on C1 and decomposing it into three features working in tandem: [±nominative], a generalized [±EPP], and [±aboutness], as shown in figure 3; we embed theses features in large brackets under the diacritic label of [±SoP] and include a V-movement feature [±V]. Thus, SoP is an epiphenomenal syntactic portmanteau. The difference between Italian and Kiezdeutsch, we argue, is simply that SoP is encoded one head higher in the latter. In SG these features simply do not conspire to create SoP effects.

Figure 3. Properties of Kiezdeutsch C¹.

We argue that the [±nominative] property of C¹ is incompatible with an accusative DP in its specifier, thus ruling out accusative object DPs with [+aboutness] but not other so-called SoPs. The nominative Case assigning property of C¹ is ultimately an innovation, assuming standardly that SG instead assigns nominative via T, following the traditional conception of nominative Case assignment (Chomsky Reference Chomsky1995, Reference Chomsky, Martin, Michaels and Uriagereka2000, Reference Chomsky and Kenstowicz2001); we return to SG explicitly below after introducing the Kiezdeutsch mechanism. Accusative Case is assigned via v to its internal DP argument, the direct object. We hold that C¹ can host a [+nominative] Case feature rather than passing it on to T via inheritance (Chomsky Reference Chomsky, Freidin, Otero and Luisa Zubizarreta2008); this follows from the position that T’s features are inherited from the phase head C in different measure, for instance, [ϕ], [+D], [+nominative] (e.g. Chomsky Reference Chomsky, Freidin, Otero and Luisa Zubizarreta2008, Ouali Reference Ouali, Roberta D’Alessandro and Hrafn Hrafnbjargarson2008, Reference Ouali2011, Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010, Miyagawa Reference Miyagawa2017). Typically, straightforward subject-related EPPs, for example, English and French, are related to both [±D] and [±nominative]. However, Spec,CP¹ is not a strict DP subject position because [+D] requirements are inherited by T, which has parameterized phrasal movement of the entire vP to Spec,TP (cf. Richards & Biberauer Reference Richards, Biberauer, den Dikken and Christina2005, Biberauer & Richards Reference Biberauer, Richards and Boeckx2006, Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010, Mohr Reference Mohr2005) to satisfy [±D] on T; this allows a head-initial TP to give the impression of being head-final (Vikner Reference Vikner2001, Haider Reference Haider2010, Walkden Reference Walkden2017), and explains why German and indeed Kiezdeutsch lack an English-style subject-EPP in Spec,TP (Abraham Reference Abraham1993, Haider Reference Haider1993, Reference Haider2010, Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010, among others).Footnote ¹⁰

Let us be explicit how this scenario blocks an accusative preverbal DP Spec,CP¹ but not in V2. Importantly, Case features are not movement triggers (Chomsky Reference Chomsky and Kenstowicz2001, Alexiadou & Anagnostopoulou Reference Alexiadou and Anagnostopoulou2001). Thus, an accusative DP can be probed by C¹’s generalized EPP and/or aboutness feature, yet incompatibility between [±nominative] on C¹ and the accusative DP is rendered only after movement, once the derivation is sent to PF. Explicitly, if Spell-Out attempts to realize an accusative DP in a nominative assigning position, the derivation will be rendered ungrammatical on account of Crash because interface conditions are prevented from fully converging (Chomsky Reference Chomsky1995:§3-4). Consequently, an accusative DP cannot remain in Spec,CP¹ and must evacuate to the higher C projection, spec,CP² in order to save the derivation, yet this movement is not driven by a probe-goal Agree relation.Footnote ¹¹ Evacuation is not forced if the accusative DP is first probed by features on C²; thus, narrow-syntactic movement would bleed the clash before Spell-Out. Therefore, an accusative [+aboutness]-bearing DP can only appear in a V2 configuration under our suggestions. In contrast, since SG’s C¹ is devoid of Case features, no Case-driven clash can occur in Spec,CP¹ and the position is totally flexible.

Our hypothesis thus raises the interesting possibility that evacuation of Spec,CP¹ by a non-D-linked accusative DP aboutness topic can plausibly be considered a case of PF-driven phrasal movement because the Case clash is a postsyntactic phenomenon, understanding the problem to emerge in Spec,C¹ when the C² Phase sends its complement to Transfer. Indeed, Chomsky (Reference Chomsky1995:358) has suggested that V2 relates to the phonological component, even claiming that much head movement may be PF-driven (Chomsky Reference Chomsky and Kenstowicz2001:37). While we do not consider V2 or head movement to be exclusively reflexes of PF (see also Roberts Reference Roberts2010), we are open to PF-driven repair operations. We note also that although we employ Case features specified for nominative and accusative, a dependent-Case approach with computationally indistinct Case features (Chomsky Reference Chomsky and Kenstowicz2001, Alexiadou & Anagnostopoulou Reference Alexiadou and Anagnostopoulou2001) is compatible with our proposals (for discussions see McFadden Reference McFadden2004, Bárány & Sheehan Reference Bárány, Sheehan, Anagnostopoulou, Sevdali and Mertyris2024). This would require that the object-DP, which is assigned syntactic Case before receiving accusative form at PF, cannot remain in the spec of a head whose Case feature associates with the unmarked Case at PF (i.e. nominative).

Moving on, we have explained why accusative DPs cannot surface in Spec,CP¹, but not why they are blocked from evacuation in V3. A preverbal accusative DP is impossible, we argue, because an XP merged in spec,CP², such as a framesetting adverbial (or innovative high central adverbs; see Sluckin & Bunk Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023, Breitbarth Reference Breitbarth2023), blocks evacuation of Spec,CP¹, thus preventing the derivation from converging. A condition is then necessary whereby preexisting PF-realization of Spec,CP² blocks such a postsyntactic repair. A solution is that satisfaction of the EPP feature on C² via (external) Merge of any XP in Spec,CP² deactivates its ability to probe downwards and build structure for moved items. This explanation allows multiple specifiers for external but not internal merge; therefore, it correctly allows stacking of adverbials and HTs in V3> (see also Sluckin & Bunk Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023) above SoP-like elements but not above others.

Finally, our proposal successfully derives bottleneck effects (cf. Haegeman Reference Haegeman1996, Roberts Reference Roberts and Rizzi2004). Decomposing SoP allows CP¹ to function as a gateway to the higher CP, when aboutness and nominative Case are valued in situ, for instance, if some different category is scrambled closer to C¹’s EPP feature, or the SoP is nested in a focused vP structure. Since incompatibility with Case only arises after movement, Spec,CP¹ cannot block movement of a syntactic category. Non-SoPs XPs which clash in Case will either be probed C² anyway and move, forced to evacuate Spec,CP¹ to Spec,C² as set out, or lead to crash. However, a preference for nominative subject DPs/pronouns arises through their tendency to encode aboutness; if an appropriate goal is a nonnominative SoP-element, such as a spatio-temporal element, a situational argument, or a dative experiencer (although largely absent in our data), they can land in Spec,CP¹ without issue, assuming that (i) they agree with C¹’s [+aboutness] feature and (ii) nominative can be assigned downwards via Agree. Since dative DPs receive inherent Case, they are not banned in Spec,CP¹, as structural Case requirements are blind to them (McFadden Reference McFadden2004, Woolford Reference Woolford2006) and there is no ambiguity with a non-nominative structural Case at PF.

7.2 Some notes on structural versus inherent Case in SG and Kiezdeutsch

An anonymous reviewer notes that examples such as (26) in SG show that nominative can be assigned low in vP (see also Sigurðsson Reference Sigurðsson, Brandner and Zinsmeister2003, McFadden Reference McFadden2004, Marantz Reference Marantz2008, Schäfer Reference Schäfer2008, Caha Reference Caha2009). They ask if Kiezdeutsch’s nominative feature on C¹ could reach down into the vP phase in such instances.

Let us first clarify why such structures are not de facto ruled out in Kiezdeutsch on grounds of phasehood; we note that the former predicate is a psych verb of the type that can be understood as 2-placed unaccusative predicate (Belletti & Rizzi Reference Belletti and Rizzi1988, McFadden Reference McFadden2004; Hirsch Reference Hirsch2018), while the second is a passive. We assume that unaccusative and passive predicates are not phasal in the same way as transitive active predicates (see the standard position in Chomsky Reference Chomsky and Kenstowicz2001). Although Chomsky (Reference Chomsky and Kenstowicz2001) says that this instance of vP is defective, this may relate to the fact that vP domain of a passive or unaccusative, in which the derived subject is always an internal argument, is not a proper predication on its own, i.e. the subject-predicate structure is ambiguous without the addition of operations/layers further up the hierarchy. Ergo, even under a dynamic phasal approach (e.g. Dikken Reference Dikken2007), the underlying structures of different types of unaccusatives and passives do not qualify as inherent phases. We would go as far as to contend that they are thus not well-formed propositions, without the addition of layers such as TP, possibly a Pass(ive)P (Alexiadou et al. Reference Alexiadou, Anagnostopoulou and Schäfer2015), or a VoiceP (Kratzer Reference Kratzer, Rooryck and Ann Zaring1996).Footnote ¹² While the nature of the vP phase is debated, it is a well-established position that examples of the type in (29) do not show phasal vPs. If correct, the only clausal phase head is C¹ and there should be no issue for nominative Case assignment in KD into a low position. However, even if the reviewer is correct that our theory rules out low nominative assignment in KD, a search of KiDKo does not find a single example of an embedded dative argument above a low vP subject of the types in (26). Indeed, we note further that we find only one instance of dative-case preservation in Kiezdeutsch passives at all and it is produced by a monolingual speaker of German (27).

Moreover, the reviewer asks why SG is less restrictive than Kiezdeutsch when it comes to logical subjects bearing accusative Case, as in (28):

Our model does not rule such structures out syntactically in Kiezdeutsch but rather only in V3, since C¹ is the position associated with Nominative Case assignment. We explain explicitly how V2 OV_FIN orders can be derived below. We note, however, that we do not find a single instance of these types of one-placed accusative predicates in Kiezdeutsch.

We reiterate that SG does not show SoP effects, which we view as a side-effect of a particular feature combination on C¹ in Kiezdeutsch. We assume that nominative Case is assigned by T to the highest visible DP in the vP-complex, which moves to Spec,TP (Richards & Biberauer Reference Richards, Biberauer, den Dikken and Christina2005, Biberauer & Richards Reference Biberauer, Richards and Boeckx2006, Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010, Mohr Reference Mohr2005). Therefore, structural nominative Case assignment is assigned lower in SG than in Kiezdeutsch and due to the pied-piped movement of vP, nominative Case appears as if it were assigned low in vP. This rases two questions: by what mechanism exactly is nominative assigned by T into the embedded vP domain? And why do we not see a similar interface-derived ban on nonnominative subjects one position lower in SG?

We suggest that there exist several options for structural nominative Case assignment to the nested DP in Spec,vP which itself sits in Spec,TP. It is achievable in several ways: (i) M-command by TP of the subject DP in spec,vP, assuming percolation of T’s features to the maximal projection; (ii) feature percolation of DP’s Case feature to the maximal projection vP, which then agrees in a simple head-spec configuration; or (iii) T’s Case feature assigns Case via C-command before vP-raising. We also note the further possibility that nominative case is truly assigned in situ by the v/Voice head itself via M-command (Schäfer Reference Schäfer2008), which if correct would force us to assume that typologically nominative Case assignment can be a parameterized capability of C, T, or v; we will not pursue this interesting option here. Thus, the more pertinent issue is, if T assigns nominative, why high accusatives appear possible in SG in contexts such as (29a), but also in a position above a nominative in embedded (29b) or Wackernagel contexts (29c).

In the case of examples such as (29), we are dealing with so-called impersonal psych-verbs, such as verlangen ‘to long for’, dürsten ‘to thirst ’, frieren ‘to freeze’, gelüsten ‘to crave’. A notable property of such impersonal psych verbs is that they can occur with overt quasi-argumental es (Cardinaletti Reference Cardinaletti1990a, Mohr Reference Mohr2005), which is often cliticized simply as ’s, a property going back to Middle High German (Lenerz Reference Lenerz and Abraham1985).Footnote ¹³ Indeed, this observation holds for a range of dative psychological predicates such as DAT+ kalt sein ‘to be cold’ or even DAT + langweilig sein. We follow Mohr (Reference Mohr2005), who assumes that these structures always involve an external quasi-argumental es which can either be overtly realised or be merged covertly as pro; indeed, the overt version is more acceptable in embedded contexts (Mohr Reference Mohr2005:186) (30).

If correct, we can explain two phenomena. Firstly, nominative case is straightforwardly assigned to the external quasi-argument of vP and we need not assume that the inherent accusative must also bear structural nominative. Secondly, the lack of agreement between apparent subjects marked with inherent accusative Case falls out naturally, because the verb is in fact simply agreeing with the quasi-argument, as is the case for weather verbs or German existential es gibt ‘there is, lit. it gives’. A further consequence of this approach, however, is that inherent accusative Case in German may not in fact be inherent at all, but is always structural. We accept that this may be a controversial side effect, yet it is perhaps desirable in providing a more succinct system overall.Footnote ¹⁵

Turning finally to the issue of what do with high accusatives in (29b,c), we simply assume for (29b) that these accusative pronouns target the relevant scrambling position that is invisible for Case assignment, such as an FP between vP and TP of the sort suggested by (Bidese Reference Bidese and Putnam2011), or higher specifiers of TP used for adjunction and thus not subject to Case effects. We now show how the proposals apply to different types of V2 and V3 with familiar topics before showing how Kiezdeutsch derives typical V2 clauses.

7.3 V3 with familiar topics

Familiar topics in V3 need not be reclassified as SoPs. Appropriate elements in Spec,CP¹ can simultaneously value aboutness, satisfy the EPP (but need not value nominative Case) on C¹ and [±Fam] requirements on C², because aboutness is a subset of Fam, namely, [+Aboutness] ⊂ [+Aboutness, +D-linking]. If an appropriate familiar topic lands in Spec,CP¹, we suggest it will freeze because either, (a) a subject DP will value all of C¹’s features in a Spec-head configuration, for example, EPP, [+Aboutness] and [+Nom]; or (b) caseless non DP-XPs, features relevant for movement enter into a spec-head Agree relationship, while the Case feature agrees downward with an appropriate DP. This is not unlike Criterial Freezing (Rizzi Reference Rizzi, Lai-Shen Cheng and Corver2006, Reference Rizzi and Phoevos Panagiotidis2010, Rizzi & Shlonsky Reference Rizzi, Shlonsky and Frascarelli2006); yet we do not employ criterial heads with single characterizing features. A local agree relation between [±Fam] on C² and the XP it C-commands in Spec,C¹ is then necessary to avoid Crash. As stated above, forced evacuation of Spec,CP¹ results from a clash in structural Case features (or their PF realization). In such instances, Spec,CP² remains available for Merge of an adverbial in V3 (pace Walkden Reference Walkden2017). An example derivation is given in figure 4. Thus, V3 is possible with SoPs and familiar topic subjects or any nonaccusative D-linked topic.

Figure 4. V3 with familiar topics in Kiezdeutsch.

7.4 Deriving OV_FIN V2 in Kiezdeutsch

As documented by Wiese & Müller (Reference Wiese, Müller, Antomo and Müller2018), Kiezdeutsch speakers of all types produce V2. We now show how multilingual Kiezdeutsch speakers can still produce V2 sentences with preverbal objects. The key observation is that fronted accusative objects in V2 and framesetters in V3 appear in complementary distribution. We have already discussed that Spec,CP¹ is capable of acting as a gateway to the C-domain, regardless of its Case properties. This is important because, on the assumption that C-heads are Phase heads (Chomsky Reference Chomsky and Kenstowicz2001, Reference Chomsky, Freidin, Otero and Luisa Zubizarreta2008), a fronted XP can neither be probed directly by C² nor move there directly by skipping C¹, as this would violate the Phase Impenetrability Condition (PIC) (Chomsky Reference Chomsky and Kenstowicz2001). Any DP that only satisfies EPP and fails to value C¹’s PF/LF-legible features, or clashes with them, will be free (or forced) to move further. We show the derivation of such an object-initial V2 clause in figure 5 produced by a confirmed multilingual speaker.

Figure 5. Derivation of object-initial V2 in Kiezdeutsch.

Firstly, the generalized EPP and aboutness features of C¹ probe together and find an appropriate goal. The DP moves to Spec,CP¹, yet this goal is an accusative DP. While the nominative Case feature on C¹ is forced to agree long range with an inverted subject, CP¹ could not be successfully spelled out because interface requirements relating to the realization of structural Case cannot converge. Consequently, the accusative DP must evacuate; yet if the DP is D-linked, it will be probed by a corresponding topic feature on C², [±Fam] in Walkden’s (2017) terms, or indeed any other relevant feature, such as [±Foc] or [±Contr], before evacuation is forced. This course of operations values all features, rules out V3 with undesirable XPs, and preserves the bottleneck. In contrast, for infrequent V3 in stricter standard-aligning spoken German (cf. Schalowski Reference Schalowski2015, Reference Schalowski2017, Bunk Reference Bunk2020), we follow Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023) who consider these a last-resort operation when speakers omit the adverb(ial) in the initial numeration (see also Haegeman & Greco Reference Haegeman and Greco2018, Greco & Haegeman Reference Greco, Haegeman, Woods and Wolfe2020), merging them late in, Spec,FrameP (Spec,CP² for our purposes).

Finally, we assumed that satisfaction of Kiezdeutsch C²’s generalized EPP feature via (external) Merge in its spec deactivates its ability to probe downwards, thus producing a complementary distribution between framesetters and accusative DPs in V3. Given instances of stacked frame adverbials (31) and stacked HTs reported by Sluckin & Bunk (Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023) leading to V4> orders, we consider that multiple specifiers on C² are a necessity. A Kaynian (Reference Kayne1994) and cartographic single specifiers requirement can also model multiple merged elements in the high CP (e.g. Benincà & Poletto Reference Benincà, Poletto and Rizzi2004, Hinterhölzl Reference Hinterhölzl2017, Wolfe Reference Wolfe2019, Sluckin & Bunk Reference Sluckin, Bunk, Karen de Clercq, Lohndal and Meklenborg2023); however, this requires recursive FrameP or HT projections above Force and does thus not make significantly different predictions. We turn to the syntax of resumption in the next section.

An updated Kiezdeutsch C-domain is given in figure 6. The full range of possible information structural categories are listed as features on their respective probing C heads. In line with the examples above, [±SoP] is illustrative with formal ingredients in large brackets.

Figure 6. The Kiezdeutsch C-domain.

This constellation overlooks operator-driven structures with wh-operators, comparatives, relatives, and equivative constructions, which we assume to occur via features/operators associated with specific clause typing C-heads (see, for example, Bacskai-Atkari Reference Bacskai-Atkari2014, Reference Bacskai-Atkari2020, Reference Bacskai-Atkari2023), a necessarily noncartographic assumption. Likewise, while their selectional properties may vary, the lexicalization of two CPs is a necessity in SG combinations such as als ob ‘as if’ or the colloquial als wie ‘than (how)’ (Bacskai-Atkari Reference Bacskai-Atkari2014, Reference Bacskai-Atkari2023). In instances affecting only the low locus of V2, namely C¹, for example certain wh-operators, framesetters are conceivably permissible above wh-items, if C² can be constant. This is a potential explanation for very rare examples of V3 in wh-contexts such as (32a). In contrast, following Walkden (Reference Walkden2017), why-type interrogatives may have a higher locus, explaining rare instances of the type in (32b). We leave the exact syntax of a specific operator-related clause types in Kiezdeutsch for further research, yet it not theoretical necessity for us that such clause types must fall out from the CP of a typical indicative matrix clause.

7.5 Applying the analysis to HTLD and LD

The proposed analysis also derives restrictions on preposed objects in Kiezdeutsch LD and HTLD. We first consider HTLD; HTs are unintegrated and externally merged above the CP locus of V2 (Grohmann Reference Grohmann2003, Frey Reference Frey, Lohnstein and Trissler2004b, Cinque Reference Cinque, Bonami and Cabredo Hofherr2008, den Dikken & Surányi Reference Dikken and Surányi2017, Haegeman & Greco Reference Haegeman and Greco2018, Wolfe Reference Wolfe2019, Samo Reference Samo2019), for example, Grohmann’s (2003) in an adjunct CP or a Force-external frame-field (Benincà & Poletto Reference Benincà, Poletto and Rizzi2004). We thus take HTs to merge in Spec,CP2, producing a verb-third order without violating the bottleneck; we assume the basic structure in (33) for SG.

Modifying (33) for Kiezdeutsch, the resumptive moves to Spec,CP¹ to value [±aboutness] and EPP requirements and the HT external merges in Spec,CP², valuing C²’s corresponding features (see Figure 7). We omit the diacritic [±SoP], referring only to its subparts. C¹’s [+nominative] feature prevents an accusative DP from being spelled out in Spec,CP¹, while the HT has no restriction against accusative objects, which is confirmed in our data.

Figure 7. Derivation of HTLD in Kiezdeutsch.

Shifting to LD, the difference between LD and HTLD is the presence of XP movement to Spec,CP2 (Grohmann Reference Grohmann1997, Reference Grohmann, Billerey and Danielle Lillehaugen2000, Reference Grohmann2003, Grewendorf Reference Grewendorf2002b,a, Reference Grewendorf, Benjamin Shaer, Frey and Maienborn2009). We adopt a big-DP analysis (cf. Grewendorf Reference Grewendorf2002a, Reference Grewendorf, Benjamin Shaer, Frey and Maienborn2009) in which both dislocate and RP merge in a complex XP (34a). This structure applies also to adverbial PPs (34b).

We assume the structure in (35) for SG LD; the big-DP moves to Spec,CP¹, the locus of the bottleneck (Spec,FinP for Grewendorf Reference Grewendorf2002a,b, Reference Grewendorf, Benjamin Shaer, Frey and Maienborn2009), yet two DPs are thus carried to the prefield. The DP (or PP/AdvP) in the specifier of the dominating DP then moves to a higher topic position.

For Kiezdeutsch, the strong aversion to preverbal accusative objects in LD and HTLD among multilinguals/interlocutors indicates the permanency of nominative assigning properties of C¹ and not simply a conventionalized property only for noncanonical V3. A Kiezdeutsch LD derivation is given in (figure 8), which we describe stepwise below. For simplicity, we do not differentiate referential topic types because the interpretation of the dislocate is not fixed (see Frey Reference Frey, Lohnstein and Trissler2004b) although often contrastive (cf. Grohmann Reference Grohmann1997, Reference Grohmann, Billerey and Danielle Lillehaugen2000).

Figure 8. Derivation of LD in Kiezdeutsch.

In figure 8, [±EPP] and [+Aboutness] probe for a lower goal, the big-DP subject whose [+Aboutness] feature is visible on DP_MAX. The big-DP moves to Spec,CP¹ where it both satisfies C¹’s generalized EPP and receives nominative Case. The maximal projection is frozen as all features of C¹ are valued in a spec-head configuration. The nested full-DP der Rest is then probed by the relevant [±Top] feature on C² and extracted to Spec,CP².

This system cannot strictly prevent a fronted accusative object big-DP from moving to Spec,CP¹ and forcibly then evacuating higher. However, the contents of the big-DP cannot then be separated; if a big-DP is not split up, the processing or information-structural advantage of resumption is lost. If a complete big-DP occupied Spec,CP², two ϕ-indistinct but informationally distinct D-elements would occupy a position that we assume can only be associated with one informational category. Moreover, if a big-DP can remain intact, ungrammatical in-situ doubling of θ and ϕ-indistinct arguments as in (36) should be possible. We thus suppose that in a complete big-DP the featurally poorer pronoun argument is subject to deletion at PF on account of its overlapping D and ϕ content, thus producing a standard V2 clause.

In sum, the following different feature specifications hold for Kiezdeutsch and SG (Table 14):

Table 14. The distribution of formal features across C¹, C², and T in Kiezdeutsch and SG

8.1 The continuity account

Some scholars consider V3 a historically consistent pattern inherited from a historical stage of the language (Wiese Reference Wiese, Abraham and Liess2013, Demske & Wiese Reference Demske and Wiese2016, Schalowski Reference Schalowski2017, Wiese & Rehbein Reference Wiese and Rehbein2016, Wiese & Müller Reference Wiese, Müller, Antomo and Müller2018) which has somehow taken root in Kiezdeutsch.

Two factors underpin this view; firstly, monolingual speakers of German appear to produce V3 rarely following the Adv→Subject→V_FIN (Schalowski Reference Schalowski2015, Reference Schalowski2017, Bunk Reference Bunk2020; Wiese et al. Reference Wiese, Öncü, Müller, Wittenberg, Wolfe and Woods2020)Footnote ¹⁶ . Secondly, framesetter-initial V3 patterns in Early New High German (ENHG) (37) (Speyer Reference Speyer2008) and Middle Low German (MLG) (38) (Petrova Reference Petrova2012) purportedly demonstrate historical continuity.

We focus on ENHG data, as neither spoken German nor the relevant Berlin dialect syntactically derive from MLG (cf. Lasch Reference Lasch1928). V3 data from ENHG is not robust enough to support historical continuity into Modern German and Kiezdeutsch. Speyer (Reference Speyer2008) finds that multiple occupation of the prefield was rare even in ENHG, 0.07 percent of his entire sample. Although Speyer (Reference Speyer2008) finds that the preverbal element was usually a subject NP, the preverbal element could be either a topic or contrastive element. The initial element could be a framesetter or contrastive element. Speyer thus posits the left-periphery in (39).

The example in (38) shows Frame > Contrast, which is not found for Kiezdeutsch, although our analysis cannot exclude a contrastive topic subject in V3. In contrast, (40a–b) show ENHG orders involving Frame > Topic (40a) and Contrast > Topic (40b) (see also Catasso Reference Catasso2021 for a review of V3> orders in Middle and Early New High German).

Moreover, although rare, O>S>V_FIN orders such as (41) were possible, which are never found in Kiezdeutsch. The lack of OSV_FIN in Kiezdeutsch and rare SG V3 suggests a different syntax.

While Kiezdeutsch shows a small subset of ENHG orders, linear similarity does not necessarily derive from systemic overlap; String similarity does not equate to grammatical continuity, especially if the V2-configuration has changed such that other historical “violations” are lost. The ENHG C-domain does not survive in modern German or Kiezdeutsch. The evidence from Kiezdeutsch V3, LD, HTLD suggests that its C-domain shows novel properties. Indeed, studies by Axel (Reference Axel2007), Speyer (Reference Speyer2008), Fuß (Reference Fuß2008), and Catasso (Reference Catasso2021), to name a few, show that no historical stage of German showed the same restrictive properties as those in Kiezdeutsch across multiple V3> orders. Moreover, the lack of any restriction on the initial constituent in SG LD and HTLD is further evidence that nominative properties were never a property of the lowest C head in the history of German.

Indeed, recently Breitbarth (Reference Breitbarth2023) has argued that sparse V3 orders with central adverbials in SG-aligning spoken language do not result from historical continuity with ENHG or MLG, supporting our position. Breitbarth (Reference Breitbarth2022, Reference Breitbarth2023) finds that such noninverted V3 in spoken German is a marginal rare option, yet demonstrates information-structural and prosodic systematicity. She also notes the relative sparsity of the pattern in TüBa-D/S (see also Sluckin Reference Sluckin2021:§7); overall V3 makes up 0.16 percent (not limited to central adverbials) of instances in TüBa-D/S (calculated from data reported by Bunk Reference Bunk2020) and 0.17 percent of the monoethnic youth language in KiDKo-mo, of which several instances are dubious (cf. Walkden Reference Walkden2017). Furthermore, Breitbarth’s (2023) experimental results show that despite low ratings, V3 is conditioned by gender and age; women under 40 are more accepting, while men were not. Breitbarth (Reference Breitbarth2023:33) thus understands the low frequency overall as evidence of a change “under the radar of social awareness.” In contrast, Kiezdeutsch V3 is a sociolinguistic stereotype.

While rare instances of V3 in colloquial SG could have contributed to its pervasiveness in Kiezdeutsch, additional factors speak against this hypothesis. Since most Kiezdeutsch speakers are not from German-speaking homes and move in social networks characterized by different types of bi/multilingualism, it is unclear if speakers had adequate exposure to such a rare pattern. Given the sparsity of V3 in Spoken German, we doubt that these instances were frequent enough in the input of children with reduced contact with L1 monolingual/monoethnic German speakers in such urban contact situations, namely, in GSLA (Cheshire et al. Reference Cheshire, Kerswill, Fox and Torgersen2011). Indeed, Breitbarth (Reference Breitbarth2023) entertains but does not commit to the possibility that monolingual speakers’ use of V3 with central adverbials could be affected by knowledge of L2 or potentially related to contact with speakers from multilingual backgrounds, noting that both the languages (Arabic, Turkish, Kurdish) and V3 associated with Kiezdeutsch speakers have existed in the German linguistic ecosystem for 50–60 years (see, for example, Barbour & Stevenson Reference Barbour and Stevenson1998, Keim Reference Keim1984). Thus, Kiezdeutsch V3 cannot be examined outside the context of a known proclivity toward SV_FIN V3 by L2 German speakers (Clahsen & Muysken Reference Clahsen and Muysken1986, Platzack Reference Platzack2001, Meisel Reference Meisel2009, Reference Meisel2011b, Walkden Reference Walkden2017), even if Kiezdeutsch speakers clearly master V2. Moreover, children exposed to German later in childhood, after the age of 4, also pass through a phase of producing more V3 (Jabnoun Reference Jabnoun2006, Rothweiler Reference Rothweiler and Lleó2006, Sopata Reference Sopata, Prior, Watanabe and Lee2010). These factors increase V3 in the linguistic ecology of multi- lingual/multiethnic speech communities. Hence, the impetus lies most plausibly in the nature of CLA in such multilingual/ethnic urban settings.

8.2 Walkden’s competing PLD account

Walkden (Reference Walkden2017) views V3 as an innovation stemming from children’s attempts to reconcile PLD from adult SVO L2 grammars lacking comprehensive V2 (Clahsen & Muysken Reference Clahsen and Muysken1986, Meisel Reference Meisel2009, Reference Meisel2011b,a, Meisel et al. Reference Meisel, Elsig and Rinke2013) and the V2 grammars of L1 monolingual German- speaking peers. We briefly summarize his position before reexamining certain aspects.

Walkden (Reference Walkden2017:69ff) proposes a three-stage progression in which acquirers must reconcile L2 and L1 derivations in (42a, b), leading to the Kiezdeutsch grammar in (42c).

These stages proceed as follows

1. i. Stage 1: L2 speakers fail to posit V-to-C and produce V3 via CP adjunction (see also Clahsen & Muysken Reference Clahsen and Muysken1986, Meisel Reference Meisel2011b). Walkden (Reference Walkden2017) considers this the source of a necessary level of nontarget-like PLD containing V3. This stage increases V3 in the German-speaking linguistic ecology (cf. Mufwene Reference Mufwene2001, Cheshire et al. Reference Cheshire, Kerswill, Fox and Torgersen2011, Wiese & Rehbein Reference Wiese and Rehbein2016) in multilingual/multiethnic communities.

2. ii. Stage 2: L1 acquisition by the children of L2 speakers. Children are exposed to adult non-V2 L2 grammars, which is interpreted as evidence for a preverbal subject requirement, presumably in Spec,TP. Yet, these children are also exposed in their peer group to L1 V2 grammars with V-to-C movement to (a single) CP. Walkden (Reference Walkden2017) argues that acquirers’ attempts to reconcile these two grammars lead them to conflate features from both and thus posit the double-CP structure (50c).

3. iii. Stage 3: Propagation of %V3 among adolescents (around 12 years of age), as it becomes a sociolinguistically salient in-group marker.

However, treating CLA as the locus of change, stage 1 simply creates the necessary conditions and stage 3 is a sociolinguistic consequence of the innovation. Kiezdeutsch speakers from monolingual German-speaking homes arguably produce stereotypical V3 as an E-language phenomenon. This prediction is supported by the fact that German monolingual Kiezdeutsch speakers more liberally employ fronted objects in LD and HTLD.

8.3 Reassessing the competing PLD account

Recall that 84 percent of the school population from which KiDKo collected data spoke home languages other than German. This matches the conditions of GSLA suggested in Cheshire et al. (Reference Cheshire, Kerswill, Fox and Torgersen2011) in which both adult L2 and L1 adults’ features tend to be rejected. This may weaken the viability of adult L2 grammars with V3 as source context; although without credible alternatives, this is not yet compelling counterargumentation. However, 84 percent is greater than the national average for second-generation Germans who do not speak German at home, which is roughly 55–60 percent (Reiss et al. Reference Reiss, Christine Sälzer, Klieme and Köller2016). Walkden’s scenario for change entails primary caregivers and children interacting in the adult L2; this is perhaps overly speculative for nonmixed families. While children undoubtedly hear their parents speaking German in public, such exposure at home cannot be readily assumed. A competing PLD account also requires considerable interaction with L1 peers at an early stage of CLA. However, we lack information on the relevant speakers’ early education; if the ethnographic make-up of preschools and elementary schools resembles the secondary schools in these areas, L1 German speakers could comprise only around 15 percent of children. If so, instead of reconciling adult L2 German and peer L1 German, a majority of children will lack German PLD at home. This means that for many their first primary exposure to German takes place later than monolingual L1 German speakers and simultaneous bilinguals. Consequently, their main German input will come from a minority of L1 peers and teachers and a majority of “very early L2” peers (see Unsworth Reference Unsworth2016), acquiring German together from one another in GLSA.

There is reason to assume (a) an ethnographic make-up in preschools in communities resembling the KiDKo population, and (b) later exposure to German via commencement of preschool. Becker (Reference Becker, Becker and Reimer2010:23) reports that around 2010 monoethnic German children tended to begin preschool before age 3, yet children from the relevant migrant backgrounds often began after 3. Crucially, the V2 parameter is already set in L1 by around 2;5 years (Penner Reference Penner and Meisel1992). Moreover, Becker (Reference Becker, Becker and Reimer2010) reports that children from a Turkish background more often attend preschools with a higher percentage of children from migrant backgrounds, mirroring the situation reported by Wiese et al. (Reference Wiese, Freywald, Schalowski and Mayr2012).This is not only the result of more nonethnic Germans living in areas where Turkish populations concentrate, but also recommendations in these social networks play an important role in preschool selection Becker (Reference Becker, Becker and Reimer2010); but only a quarter of parents from Turkish backgrounds in Becker’s study reported having monoethnic German friends. This means members of the same social network send their children disproportionately to the same local preschools. These factors conspire to amplify the concentration of children from non-German-speaking backgrounds in preschools beyond the demographics of multiethnic areas.

In sum, many children from the relevant communities are likely to be exposed to German later and are disproportionately concentrated in environments with reduced L1 sources of German PLD. Biedinger & Becker (Reference Biedinger, Becker, Becker and Reimer2010) found that children from Turkish-speaking homes who attended preschool longer performed better in German once school age is reached; notably, children from preschools with higher concentrations of multiethnic children were more likely to require extra support in German at school. This is unsurprising given the reduced L1 input from peers in situations of GSLA (section 2). But how could these factors influence the emergence of V3?

9.1 Different bilingual types and associated effects

Bilinguals can be divided into three groups according to the age of onset, that is, age of first exposure to a language (Tsimpli Reference Tsimpli2014:284):

1. i. Simultaneous Bilinguals (exposure to two or more languages from birth)

2. ii. Early Successive Bilinguals (exposure before age 4)

3. iii. Late Successive Bilinguals (exposure from age 4–8)Footnote ¹⁷

Later exposure is associated with onset effects and interacts with general maturational effects in CLA (Tsimpli Reference Tsimpli2014). Building on observations from the Interface Hypothesis (see Sorace & Serratrice Reference Sorace and Serratrice2009, Sorace Reference Sorace2011) that the syntax–pragmatics interface presents challenges for bilingual acquirers, such as nontarget use of null subjects, Tsimpli (Reference Tsimpli2014) proposes that different parameters are acquired either early, late, or very late in L1 CLA. For example, narrow-syntactic and semantically vacuous (macro)parameters are acquired before those with interface conditions, for example, OV/VO before V2. Thus, structures with higher interface requirements take longer to acquire. Consequently, the syntax-discourse/pragmatics interface is more vulnerable in the grammars of successive bilinguals. Indeed, while simultaneous bilinguals play catch-up only in obtaining the same amount of PLD as monolinguals (Tsimpli Reference Tsimpli2014), early and late successive bilinguals also have to catch up on the PLD for parameters typically acquired in earlier stages of acquisition.

For V2, simultaneous and early successive bilinguals match monolingual speakers for verb placement and finiteness; in L1 this occurs early around age 2 (Meisel Reference Meisel2011b, Tsimpli Reference Tsimpli2014). However, late subsequent bilinguals reportedly produce more V3 for longer; Kroffke & Rothweiler (Reference Kroffke, Rothweiler and Vliegen2006) find that late successive German bilinguals with L1 Turkish (onset around 6;0) produce more V3 and uninflected forms in V2 positions. Sopata (Reference Sopata, Prior, Watanabe and Lee2010) finds increased V3 for L2 German speakers with Polish L1 (onset between 3;8 and 4;7) although it reduced over time. Finally, Jabnoun (Reference Jabnoun2006), found that late subsequent bilinguals produce Frame-SV_FIN, even reporting V3 with a temporal adverb followed by a preverbal locative (43) which is derivable via a low-CP SoP requirement.

Yet, if increased V3 among late successive bilinguals is typically a phase lasting until V2 is fully acquired (Meisel Reference Meisel2011b:214), why do Kiezdeutsch adolescents retain V3 after successfully acquiring V2? We draw again on Tsimpli (Reference Tsimpli2014), who argues that input effects associated with quality/unambiguity of the input influence the duration needed to acquire a parameter. Input effects can mediate maturational effects. Thus, when the clarity or quantity of input is reduced, later acquisition of phenomena with higher interface requirements is predicated. Where onset is delayed, the richness of input is especially important for successive bilinguals to catch up. The combination of later onset and reduced or more ambiguous input received in challenging linguistic settings, such as a language-contact scenario, could facilitate the nontarget-like acquisition of given phenomena. This is especially so if cut-off points for totally target-like acquisition exist between ages 4 and 7 (cf. Johnson & Newport Reference Johnson and Newport1989, Meisel Reference Meisel2011b). Put simply, the late successive bilingual child first encounters a parameter already acquired by similarly aged monolinguals and simultaneous bilinguals and must “catch up”; however, input and onset effects can conceivably combine in super-diverse multilingual/ethnic speech-communities. This child may thus never posit the same settings as monolinguals or simultaneous bilinguals, leading to an innovative grammar. Applied to V3, the input received by Kiezdeutsch speakers may have come too late and too ambiguously to rule out verb-third orders, which appear to be an emergent property (Wiese et al. Reference Wiese, Öncü, Müller, Wittenberg, Wolfe and Woods2020), while simultaneously sufficing to acquire a basic but more flexible V2 grammar; we expand on this in section 9.3.

While we lack direct evidence, some Kiezdeutsch speakers in KiDKo do show effects associated with later successive bilingualism in V3 contexts. For instance, Meisel (Reference Meisel2009, Reference Meisel2011b) and Meisel et al. (Reference Meisel, Elsig and Rinke2013) argue that exposure to German after 4;0 can result in some nontarget uses of morphosyntactic inflection, such as case, gender, and other ϕ-related agreement marking (see also Tsimpli & Hulk Reference Tsimpli and Hulk2013, Tsimpli Reference Tsimpli2014, Unsworth et al. Reference Unsworth, Frose Argyri, Hulk, Sorace and Tsimpli2014, Unsworth Reference Unsworth2016). We find several errors in gender-marking (44), finiteness inflection (45), and C/case morphology (46).

We also find rare examples indicating competition between VO/OV, for example, a postverbal particle after a nonfinite form (47), which is typically diagnostic of a left-headed VP (Fuß Reference Fuß, Ferraresi, Jäger and Weiß2018).

9.2 The information-structural primitive behind V3

We have argued that multilingual Kiezdeutsch speakers acquired V2 later and in more difficult conditions than monolingual German speakers. However, this does fully explain the resilience of V3, which emerges repeatedly in the speech of L2 adults (Clahsen & Muysken Reference Clahsen and Muysken1986), late successive bilinguals (Rothweiler Reference Rothweiler and Lleó2006, Jabnoun Reference Jabnoun2006, Sopata Reference Sopata, Prior, Watanabe and Lee2010), Heritage speakers of V2 varieties (Alexiadou & Lohndal Reference Alexiadou, Lohndal, Antomo and Müller2018), speakers of Kiezdeutsch and similar urban vernaculars (Wiese Reference Wiese, Abraham and Liess2013, Walkden Reference Walkden2017, Meelen et al. Reference Meelen, Mourigh, Lai-Shen Cheng, András Bárány, Douglas and Vikner2020), early L1 CLA (Tracy Reference Tracy1991, Platzack Reference Platzack2001, Tracy & Thoma Reference Tracy, Thoma, Dimroth and Jordens2009), and also in synchronic and historic V2 varieties which allow V3 micro-variationally (Haegeman & Greco Reference Haegeman and Greco2018, Hinterhölzl Reference Hinterhölzl2017, Wolfe Reference Wolfe2019). As noted in section 8.1, infrequent examples are even found in spoken German (Bunk Reference Bunk2020, Breitbarth Reference Breitbarth2022, Reference Breitbarth2023). Consequently, the similarity of V3 across different linguistic scenarios is beyond coincidental and poses the question as to why such orders seem to emerge repeatedly.

Indeed, Wiese et al. (Reference Wiese, Öncü, Müller, Wittenberg, Wolfe and Woods2020) on English, German, and Turkish, argue that V3 constitutes a natural order of information distinct from language-specific grammatical constraints. Where situations can be elicited which weaken normal grammatical constraints, a preference emerges for Framesetter > Subject_TOPIC > V_FIN. Although unproblematic in English, V3 and its specific discourse requirements is noncanonical in German and Turkish. Using experimental methodology, Wiese et al. (Reference Wiese, Öncü, Müller, Wittenberg, Wolfe and Woods2020) find a preference for V3-ordering of information in nonverbal tasks, for example, an experiment involving comics and the placement of items, such as clocks, verb cards, and Playmobil. However, in a verbal task, German speakers resorted to V2. This purportedly shows V3 to be a basic natural order irrespective of V2. If Wiese et al. (Reference Wiese, Öncü, Müller, Wittenberg, Wolfe and Woods2020) are correct, V3 equates to a pragmatic primitive. We suggest that such primitives can drive acquisitional biases; thus, the acquirer of a strict V2 language, like German, must posit a syntactic configuration excluding V3 during parametrization. This predicts V3 in early CLA, which German and Swedish (Tracy Reference Tracy1991, Platzack Reference Platzack2001,Tracy & Thoma Reference Tracy, Thoma, Dimroth and Jordens2009) speakers under age 3 produce; for instance, Platzack (Reference Platzack2001) notes a Swedish example (age 2;2) with an initial locative adverb där ‘there’ (48).

Since V3 encodes a particular information-structural configuration, it involves the syntax–discourse/pragmatics interface. Thus, if the syntax–pragmatics interface is particularly vulnerable in bilingual acquisition and production (Sorace & Serratrice Reference Sorace and Serratrice2009, Sorace Reference Sorace2011, Tsimpli Reference Tsimpli2014), bilingual acquisition characterized by onset or/and input effects may well resort to “natural orders” of information more readily. From a language-change perspective, we propose that Kiezdeutsch V3 results from the parametrization of this default into the syntax. We now address how this takes place.

9.3 Minimal Defaults and Phase Heads: from default to novel syntax

The C-domain is central to the interfaces between syntax, pragmatics/discourse and semantics (Rizzi Reference Rizzi and Haegeman1997, Platzack Reference Platzack2001, Benincà & Poletto Reference Benincà, Poletto and Rizzi2004, Frascarelli & Hinterhölzl Reference Frascarelli, Hinterhölzl, Schwabe and Winkler2007, Miyagawa Reference Miyagawa2017, among others). Assuming that information structure is encoded syntactically, one must ask how natural orders of information are encoded as discourse/syntactic primitives and how they drive acquisition. Here, we propose an unfurling of discrete syntactic and information-structural settings in the C and T domains, presenting a middle ground between positions assuming early full competence in C/T (cf. Poeppel & Wexler Reference Poeppel and Wexler1993), that is, all abstract principles of UG are accessible during CLA; and truncation of clausal layers (Rizzi Reference Rizzi1993, Haegeman Reference Haegeman1995).

We suggest that a Minimal Default Grammar (Roeper Reference Roeper1999:173) (MDG) available to the child acquiring language encodes a proclivity for V3 and preverbal subjects before the V2 parameter is fully acquired, that is, a more basic structure containing less structure, and, if Move is costly (see Westergaard Reference Westergaard2009), less or no movement. Before proposing an exact structure, we draw on work on null-subjects and the acquisition of topic-drop in V2 languages, which provides context for MDG-defined subject and topic positions.

Firstly, adult German and Dutch permit V1 with dropped topic objects, as in (49) (cf. Cardinaletti Reference Cardinaletti, Mascaró and Nespor1990b, Rizzi Reference Rizzi, Hoekstra and Schwartz1994, Müller & Hulk Reference Müller and Hulk2001).

However, children demonstrate a stage of nontarget-like topic/argument drop of both subjects and objects (see 50a,b), without obligatory V1, regardless of whether their L1 is a null or overt subject language (Rizzi Reference Rizzi, Hoekstra and Schwartz1994, Hamann Reference Hamann1996, Müller & Hulk Reference Müller and Hulk2001). This, in part, relates to a tendency for all children to (over)produce null-subjects (Rizzi Reference Rizzi1986, Reference Rizzi, Hoekstra and Schwartz1994). Müller & Hulk (Reference Müller and Hulk2001) argue that here children are utilizing universal pragmatic strategies in early CLA before fully acquiring the correct language-specific rules.

Based on mono- and bilingual German, Dutch, French, and Italian child data, Müller & Hulk (Reference Müller and Hulk2001) propose that nontarget-like argument drop derives from an MDG resembling (51).Footnote ¹⁸ They find that Romance–German/Dutch bilingual children stay in this phase longer, arguing that fronted object clitics with dropped DP-objects in Romance provide indirect evidence of deviant topic-drop in German/Dutch. While community languages relevant for Kiezdeutsch do not behave like Romance, the important point is that the acquisition of the syntax–discourse interface appears disruptable by either other languages in speakers’ repertoire Müller & Hulk (Reference Müller and Hulk2001) or a combination of onset and input factors (Tsimpli Reference Tsimpli2014, Unsworth et al. Reference Unsworth, Frose Argyri, Hulk, Sorace and Tsimpli2014).

While Müller & Hulk (Reference Müller and Hulk2001) omit a CP layer, we posit a maximally underspecified C-head for which the full set of features is yet to be posited, in line with Poeppel & Wexler’s (1993) Full Competence Hypothesis. Moreover, Rizzi (Reference Rizzi, Hoekstra and Schwartz1994) argues that a UG-provided strategy in CLA includes null subjects in the root specifier, namely, Spec,CP. That is, CP can be thought of as a default subject position for covert subjects in CLA. Consider also that T’s formal feature composition is ultimately inherited from C and that both C and T probe together (Chomsky Reference Chomsky, Freidin, Otero and Luisa Zubizarreta2008, Ouali Reference Ouali, Roberta D’Alessandro and Hrafn Hrafnbjargarson2008, Reference Ouali2011, Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010). If correct, the child acquiring language must tease apart T and C, rather than posit the phase head C, the source of T’s properties, after acquiring T. We thus update the MDG in (51) to (52), a collapsed C/T with a lower and higher specifier for which the full range of features and inheritance relations are yet to be posited.Footnote ¹⁹ This mirrors early suggestions by (Rizzi Reference Rizzi, Mascaró and Nespor1990) that C bears both [±C] and [±I] categorial features which, depending on the language type, can be set either positively or negatively; CP in a V2 language is set [+C, +I]. We omit a finite lexical verb on C/T in the MDG, as V raising must be acquired, as evidenced by its absence in English (Pollock Reference Pollock1989), Mainland Scandinavian (Holmberg & Platzack Reference Holmberg and Platzack1995, Vikner Reference Vikner1995), and German (Vikner Reference Vikner2005,Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010), yet Merge of auxiliaries is an inherent property of C/T.

Consequently, T’s EPP properties and the default null-subject position in Spec,CP are collapsed in the lower specifier of C/T, in line with Rizzi’s (1994) UG-provided position. Likewise, if T is the typical nominative assigning head, [±nominative] also resides on C/T. Tailoring the MDG in (52) for V3, we suggest that the higher specifier of C/T is a general position which can host not only null-objects but also adverbial adjuncts such as framesetters. Thus, the earliest instances of V3, which also appear in early monolingual grammars (see, for example, Platzack Reference Platzack2001, Tracy Reference Tracy1991), derive from the structure in (53); this produces the described complementary distribution between fronted framesetters and object DPs.

Nonetheless, children acquiring Kiezdeutsch and SG must decompose C/T into C and T and also separate the lower and higher specifiers of C/T into C¹ and C², so as to derive either both subject position effects and the V2 system in Kiezdeutsch or just the stricter V2 system in SG. When the child receives enough input to distinguish CP and TP, they will redistribute subject-associated properties of the lower Spec,C/T (i.e. [±EPP], [±Aboutness] [±D] [±nominative]) to either C or T, which depending on the feature distribution can also remove a derived subject position if evidence in the input is absent. For example, since Spec,TP in German does not host DP subjects (Abraham Reference Abraham1993, Haider Reference Haider1993, Reference Haider, Alexiadou and Hall1997, Reference Haider2010, Richards & Biberauer Reference Richards, Biberauer, den Dikken and Christina2005, Biberauer & Richards Reference Biberauer, Richards and Boeckx2006, Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010), the child acquiring SG abandons the MDG subject position entirely for the adult grammar. Thus, monolinguals and simultaneous and early successive bilinguals abandon a subject relation in the C-domain for an informationally flexible position by positing [±nominative] on T and most other discourse features on C¹. However, for Kiezdeutsch, we argue that subject-associated SoP properties are redistributed so that Spec,CP¹ retains a loose association with subjecthood, namely, [±EPP], [±Aboutness] [±nominative] on C, as proposed in section 7. Thus, if the MDG underlying both V3 and nontarget argument drop in CLA is correct, the Kiezdeutsch innovation can be summed up as containing the following parts:

1. i. The failure to posit inheritance of [±NOMINATIVE] from C to T so that Case assignment remains a property of the phase head C¹.

2. ii. The positing of a [±Aboutness] on C¹.

3. iii. The generalization of all other information-structural features associated with D-linked top ics, foci, and contrastive elements to the higher C², largely following Walkden (Reference Walkden2017).

Recall our suggestions that (i) a link exists between V3 and delayed bilingual CLA, and (ii) many Kiezdeutsch speakers are likely to be successive bilinguals with reduced early exposure to L1 German input leading to input and onset effects. We thus propose that Kiezdeutsch speakers lacked adequate evidence to entirely abandon the MDG-endowed V3. Considering the interaction between V2/3 and the syntax–pragmatics interface in the C-domain, onset and input effects may disrupt the typical course of acquisition as speakers attempt to map universal pragmatic strategies onto language-specific rules (Müller & Hulk Reference Müller and Hulk2001, Platzack Reference Platzack2001, Sorace Reference Sorace2011). We propose that the SoP requirement was innovated when sufficient numbers of speakers lacked clear counterevidence against a default MDG-driven linearization. While the MDG is not simply adopted, it is incorporated into a coherent innovative syntax, as established by positing the MDG root-specifier subject position as a more general SoP position hosting both [±Aboutness] and [±nominative]

We assume that the split C/T-split occurs when children posit inheritance of ϕ, D from C-to-T (and for SG also Case features). This might coincide with full acquisition of V-final embedded clauses with finite auxiliaries, as an auxiliary in the left-headed MDG C/T without parameterized vP-to-Spec,TP pied-piping (Mohr Reference Mohr2005, Richards & Biberauer Reference Richards, Biberauer, den Dikken and Christina2005, Biberauer & Richards Reference Biberauer, Richards and Boeckx2006) would produce an AUX_FIN>V_LEXICAL preference in embedded contexts. Indeed, evidence supporting an early combined C/T projection comes from Swiss German embedded AUX_FIN>V orders in early child datan (Gawlitzek-Maiwald et al. Reference Gawlitzek-Maiwald, Tracy, Fritzenschaft and Meisel1992, Schönenberger Reference Schönenberger2001) (54) before V>AUX_FIN is fully acquired. Although such early phenomena could result from V-to-T movement (see, for example, Westergaard et al. Reference Westergaard, Lohndal, Alexiadou, Ramshøj Christensen, Wood and Jørgensen2019), an unfurling analysis accounts for this behavior without requiring children to later unlearn an operation absent in adult German (Vikner Reference Vikner2005, Biberauer & Roberts Reference Biberauer, Roberts, Theresa Biberauer, Roberts and Sheehan2010).

We cannot definitively say when C¹ and C² split into different projections, but for Kiezdeutsch this must at least coincide with full acquisition of other V2 patterns. While unambiguous V2 input provides evidence that the C-domain may host any XP, children in the relevant multiethnic communities will lack evidence to abandon MDG-endowed high framesetters in the upper Spec,C/T. Kiezdeutsch speakers interpret the complementary distribution of high framesetters and accusative DPs, but not subjects or potential SoPs, as evidence that Spec,CP² is an A’- position where a framesetter can merge. For SG speakers, this split may be later, as we argue that the higher projection is only really important for HTLD, LD, and perhaps CP-expletive es. We consequently posit the following unfurling of structure during CLA from a single C/T projection to a two CPs and a TP projection:

Finally, we have not explicitly discussed the acquisition of rare V3 in SG beyond Sluckin & Bunk’s (2023) proposed repair strategy. If incipient SG V3 is evidence for a genuine V3 option in German (cf. Breitbarth Reference Breitbarth2023), then a possible point of microvariation is the generation point of high framesetters and adverbials; explicitly, in line with Breitbarth’s suggestions, this could be a high C-domain position. In our terms, SG speakers with this option can merge these items in spec,CP², yet they do not share the SoP-deriving aboutness and nominative properties on C¹; a preference for subjects in incipient SG V3 must thus necessarily derive via interface effects beyond the narrow syntax (for instance, in the sense of Wiese et al. Reference Wiese, Öncü, Müller, Wittenberg, Wolfe and Woods2020), yet they cannot be ruled out.