2. Properties of Slf coordination

This section begins by summarising the structural properties of SLF coordination, including the varying word order in the conjuncts, omission in the second conjunct, scope of adverbs and asymmetric extraction. Following that, the interpretation of SLF coordination will be discussed.

2.1. Structural properties of SLF coordination

An obvious trait of SLF coordination is the different word order of the two conjuncts. German is known for a V2 word order in non-embedded declarative sentences. While the first conjunct has a V2 word order as expected, the prefield (Vorfeld), i.e. the clause-initial position for the element before the finite verb, in the second conjunct must remain empty. The second conjunct must have the V1 word order, and no constituent may precede the finite verb if there is a subject gap.

Second, in SLF coordination, no constituent in the last conjunct other than the subject can be omitted. (7-a) is a grammatical example of SLF coordination in which the second conjunct contains no subject. However, omitting a further constituent (7-b) or multiple constituents (7-c) is ungrammatical.

Third, adverbial constituents in the first conjunct can apply to the whole sentence or only to the first conjunct, but not exclusively to the second conjunct. In the above (7-a), the adverbial constituent morgen ‘tomorrow’ takes the scope of the whole coordinated sentence. In (8-a), the temporal adverbial in the prefield of the first conjunct applies only to the first conjunct, for the temporal adjunct in the second conjunct indicates a different time period. However, (8-b) is not acceptable since the adverbial constituent in the prefield is intended to apply to the non-initial conjunct only. Cases like (8-b) are ruled out in SLF coordination.

This holds even for cases where the adverbial constituent is not in the prefield. In the following example (9), the adverb wohl ‘probably’ is embedded in the first conjunct, but it can still apply to both conjuncts, leading to the sole interpretation that it was probably the case that both events took place.

Fourth, although extraction from the second conjunct is not permitted, extraction from the first conjunct is possible, which seems to be violating the Coordinate Structure Constraint (CSC). The CSC states that no conjunct or element in a conjunct can be moved out of the coordinate structure (Ross, Reference Ross1967). In (10-a), the fronted constituent das Gepäck ‘the luggage’ can be considered as a constituent extracted from the first conjunct, but not as an extracted constituent from the second conjunct (10-b).

These features show an asymmetry between the conjuncts. The conjuncts may have different word order, adverbials take different scope and extraction out of coordination is also asymmetric. These observations are not in accordance with the usual assumption of symmetry for coordination.

2.2. Interpreting SLF coordination

Höhle (1983) already noted that SLF coordination is always associated with a fused interpretation which assumes a direct, natural link between the events described in the conjuncts. This can be illustrated by the following contrast:

In (11-a), there is a natural link between the two conjuncts, and the preferred interpretation is that the hope is not to be reported for being seen, while (11-b) rather refers to two disjoint events, i.e. the speaker hopes to be neither seen nor reported (Höhle, Reference Höhle, Müller, Reis and Richter1983). It should be noted that the subjects in the above sentences involve the use of a negative quantifier. For cases where an impersonal pronoun as in (6) or a proper name as in (7) is used, repeating the subject overtly in the second conjunct does not seem to give rise to another interpretation. Instead, the version with the overt subject in the second conjunct only sounds marked or less natural, which is illustrated in the following examples:

Based on these observations, we can construct structures that are similar to (11) to illustrate how a subject gap can lead to differences in semantic representations. The following pair of sentences shows a comparison in terms of truth-conditional meaning between SLF coordination and phrasal coordination. In (14-a), it should be the same individual who both saw and reported the speakers, and its meaning is represented in (14-b), but in (14-c), the sentence concerns two different individuals, as is shown in (14-d), due to the use of a repeated indefinite NP which introduces a new referent.

Reich (Reference Reich, Fabricius-Hansen and Ramm2008) proposed that in SLF coordination, it is obligatory to interpret the sentence as describing a single event instead of two separate events. Therefore, unrelated events cannot be described in SLF coordination (Weisser, Reference Weisser2019)

At this point, it should be noted that although the above example taken from Weisser (Reference Weisser2019) was judged as ungrammatical, marked by an asterisk, it should be better judged as peculiar or inappropriate in terms of meaning, thus with a hash mark instead. The oddness of the sentence is not due to ungrammaticality but should be attributed to pragmatic infelicity. This issue will be discussed in more detail later in Section 4.2.

3. A critical appraisal of existing theoretical approaches

SLF coordination has received broad attention and different theoretical accounts have been developed addressing its form and meaning. This section summarises current theoretical approaches to SLF coordination. These approaches can be divided into two categories: (1) coordination analyses, which assume SLF coordination is a case of coordinate structure with special syntactic properties; (2) adjunct analyses, which treat the second conjunct without a subject as subordinate to the first conjunct.

3.1. Coordination analyses

The first group of approaches treat SLF coordination as a coordinate structure and attempt to derive its properties on the basis of the general pattern of coordination. These approaches offer various solutions to the licensing of subject gaps in the non-initial conjunct, aiming to explain the special structural properties of SLF coordination, including the V1 order and the empty prefield of the second conjunct. They can be further categorised into two subgroups. Some argue that in an SLF coordination, two full clauses are conjoined, and the derivation involves ellipsis in the second conjunct, while others assume that instead of two full clauses, phrases smaller than CPs are conjoined.

3.1.1. Coordination of large conjuncts

Wilder (Reference Wilder1994, Reference Wilder, Alexiadou and Hall1997) proposed that SLF coordination is an instance of normal clausal coordination in which two full CPs are conjoined, and the subject in the second conjunct is elided phonologically due to its identity with the corresponding constituent in the first conjunct. Therefore, an SLF construction will have the following syntactic structure:

Following this approach, no extraction out of the coordinate structure is involved and SLF coordination is considered as derived from the common coordination. The deletion of the subject in the second conjunct takes place at the phonological level.

Similarly, te Velde (Reference te Velde2006) also put forward an approach to SLF coordination that involves ellipsis. Although te Velde proposed a different model for coordinate structures, in which the second conjunct is deeply embedded at the right branch of the first conjunct, it is also assumed that SLF coordination shares a similar structure with the coordination of two clauses, and the subject in the non-initial conjunct is deleted at the phonological level. Within this account, there are two major assumptions that are critical to the derivation of SLF coordination. The first one is about the general syntactic structure of Germanic languages, which states that subject-initial V2 declarative sentences are TPs, and other V2 declaratives are derived at a higher CP level. Therefore, the first conjunct is derived at the CP level, since it is an object-initial V2 declarative, with an object at the SpecCP position. If a subject-initial V2 sentence is coordinated with a sentence of a different word order, the conjoined sentence has the following structure, where the second conjunct is embedded under a node within the first TP.

The other major assumption is about ellipsis in coordination: The coordinator can license a left-edge gap in the SpecTP or the SpecCP position if the corresponding constituents are in parallel positions. In the above example (17), the element Karl in the second conjunct is a duplicate of an identical element in the first one, and both constituents occur in the SpecTP position, so the deletion of duplicate phonetic features in the second conjunct is licensed by the clausal coordinator, resulting in a subject gap in the second clause.

However, the large conjunct and ellipsis approaches face the common empirical problem that restoring the omitted constituents can lead to a different interpretation. This hints at a systematic difference between SLF constructions and their apparent counterparts without deletion operation. It applies both to normal phrasal coordination (19) as well as to SLF coordination (20-a).

In (19-a), the sentence is interpreted as describing two actions performed by the same person, while this is not necessarily the case for (19-b), which is related to two different individuals. The same applies to SLF coordination. A sentence like (20-a) does not have the same meaning as in (20-b), the version with the omitted subject restored. Instead, the quantified subject in (20-a) must take a wide scope over the whole coordinate structure, resulting in the interpretation that there are many bombers for whom both situations in the conjoined predicate apply (Sternefeld, Reference Sternefeld2006: 599).Footnote ² Such examples do not provide support for analysing the subject gap as a phonologically elided element. This contrasts with Right Node Raising (RNR), where the predominant interpretation of a quantified NP in a sentence with an ellipsis in the first conjunct (21-a) is still distributed, just as in a sentence with no ellipsis (21-b), so RNR can be analysed as a phonological reduction (Hartmann, Reference Hartmann2000). Unlike the examples in (19), for both sentences in (21), two different novels are involved.

3.1.2. Coordination of small conjuncts

In the other subgroup of coordination approaches, it is assumed that instead of two full clauses, two smaller conjuncts are coordinated in SLF coordination and therefore no ellipsis is involved in the derivation process. This type of analysis can be traced back as far as Höhle (Reference Höhle, Mascaró and Nespor1990), who proposed that the SLF coordination is the coordination of a V projection, here labelled as V’, and a functional head projection, labelled as I’. In the second conjunct, the prefield is occupied by an empty element e, which is assigned to a position in the VP by the I’ projection, but the empty element is neither case-marked nor theta-marked, leaving a subject gap in the second conjunct.

Similarly, Heycock & Kroch (Reference Heycock and Kroch1994) also proposed that in an SLF coordination, conjuncts of different categories are conjoined. A prerequisite of their account is that if the head I and all other materials at the IP level have been moved out of it, their traces should be deleted and the IP will vanish, resulting in a mixed structure CP/IP. This is exactly the case of a subject-initial V2 sentence as (23-a), in which all elements have left the IP layer and the CP has taken over all functions of the IP, so the sentence should be analysed as a CP/IP as in (23-b).

In (23-b), the intermediate projection C’/I’ accounts for the V2 word order due to properties of a C’, and it can also be coordinated with another I’ because it has inherited the traits of an I’ as well. Therefore, SLF coordination is a coordination of an I’ and a mixed projection C’/I’:

Another theory which postulates the coordination of two small conjuncts was proposed by Johnson (Reference Johnson2002). In this approach, it is assumed that there is a functional projection for verb projection raising from the right edge to a fronted position in subordinate clauses in certain Germanic languages, e.g. West Flemish as in (25), and this functional projection between the VP and the IP level is labelled as FP.

Johnson proposed the following derivation process for SLF coordination based on the assumption about FP and the observation of the CSC: First, two F’ are coordinated. Next, the first conjunct is moved out of the coordinate structure to the SpecFP position, which is specifically assumed to be free from the CSC. Finally, the finite verb and the subject are moved out of the SpecFP to CP level.

There are two major concerns, one conceptual and one empirical, specifically about the assumptions in this proposal. The first one is that the application of the CSC seems to be highly variable. Johnson (Reference Johnson2002: 118) assumed the CSC to hold for coordinate structures, but then specified that it does not apply to element extraction out of coordination of FPs. The other issue is that, unlike West Flemish, there does not seem to be a subordinating structure similar to (25) in German,Footnote ³ making the existence of an FP projection in German arguable.

A common problem faced by all three approaches, as Mayr & Schmitt (Reference Mayr, Schmitt, Everaert and van Riemsdijk2017) pointed out, is that the empirical data do not seem to support an analysis assuming coordination at a level lower than the C’. As the following example (28) shows, it is ungrammatical to introduce an additional IP conjunct into a coordinate structure that involves SLF:

3.2. Adjunct analyses

The second group of approaches do not agree that SLF coordination can be derived from ordinary coordinate structures. They analyse the second conjunct as an adjunct that subordinates to the first one, and the subject in the second conjunct can be either pronominalised or left empty.

In observation of many subordination effects involved in SLF coordination, including the scope of negation and binding, Büring & Hartmann (Reference Büring and Hartmann1998) suggested an adjunct analysis, in which the second conjunct is considered as a full CP that adjoins to the first conjunct. Within the second conjunct, the SpecCP position is occupied by an empty operator Op, which is bounded by the subject in the first conjunct, and this Op binds in turn another empty element e in the SpecIP position of the second conjunct.

This analysis provides an explanation for the subordination effects in the SLF coordination. One remaining empirical question is that within this approach, an object gap should also be licensed in the same way as a subject gap is. But an object gap does not seem to be possible in a similar construction that nevertheless allows a subject gap.

Following this line of reasoning, Reich (Reference Reich, Fabricius-Hansen and Ramm2008, Reference Reich2009) proposed another adjunct analysis in which the second conjunct is a functional projection OccP, marking a fused interpretation. The head of an OccP licenses an empty element pro in the SpecOccP position. The structure can be schematically illustrated as follows.

Despite the use of the notion pro, this pro element in (31) is not pronominal, but an empty element which only bears a [nom] feature, marking the nominative case of this element. The reason for making this assumption is that, on the one hand, assuming that pro is pronominal would cause empirical difficulties, because examples with explicit pronouns like (32) indicate that anaphoric relations cannot be established.

On the other hand, pro must carry a [nom] case feature in order to eliminate the possibility of an object gap. Although the assumptions about a unique OccP and an empty pro are not very usual, an adjunct analysis provides reasonable explanations for the properties of SLF coordination in terms of structure and interpretation.

A more recent approach was raised by Weisser (Reference Weisser2019), who assumes that SLF coordination has an underlying subordinate structure. Following this account, the second conjunct is first generated as a FinP adjunct to the VP in the first conjunct (the matrix sentence). The clause-initial position in the FinP adjunct is a PRO, bounded by the subject in the matrix sentence. If the first conjunct is constructed as a full FinP, the coordination takes place, and the second conjunct is linearised to the Spec&P position to the right of the & head to achieve consistency with the chronological order of the events. In the last step, a constituent is extracted asymmetrically from the first conjunct to the prefield. The derivation can be schematically illustrated in a simplified manner as follows:

A noteworthy point about this account is that the subject gap is regarded as a PRO, despite the usual assumption on its distribution that a PRO occurs only in non-finite sentences.

As noted by Altshuler & Truswell (Reference Altshuler and Truswell2022), treating the second conjunct as an adjunct is very similar to the model of coordinate structures proposed by Munn (Reference Munn1993). In this binary branched model, a non-initial conjunct in a coordinate structure is treated as an adjunct that adjoins the initial conjunct, and only extraction from the initial conjunct is allowed, since it is not an island. The model of Munn (Reference Munn1993), which is based on asymmetric assumptions about the coordinate structure, is very promising in providing a unified analysis of SLF coordination and other coordinate structures. The two coordinate structures will be compared in Section 4.

4. Comparing SLF Coordination and Phrasal Coordination

The current study aims to show that SLF coordination is not fundamentally different from a normal coordinate structure and can be explained by the model for coordination proposed by Munn (Reference Munn1993). This section first compares SLF coordination and the so-called ‘normal’ coordination of two clauses with no subject gaps before we move on to the analysis. Specifically, there are two possible realisations of clausal coordination without subject gaps. One possibility is that each clause contains a subject, possibly with an anaphorical relation, resulting in a coordination of two full clauses. In clausal coordination, the conjuncts are of the same syntactic status, as the subject has been overtly realised in both of them. Each conjunct can stand alone as a separate, independent sentence.

The other possibility is that the two clauses share the same subject at the beginning of the coordinate structure. In (35), there is only one single subject er ‘he’ in the sentence-initial position.

The question associated with such cases is whether they should be analysed as a phrasal coordination of two small conjuncts (36-a), or as a clausal coordination of two large conjuncts with an ellipsis (36-b) (Hartmann, Reference Hartmann, Kiss and Alexiadou2015).

If it is assumed that the coordinands are two full clauses, the subject of the second conjunct should then be regarded as omitted (Wilder, Reference Wilder1994). In terms of meaning, it seems that the analysis with large conjuncts and ellipsis is less preferable, since reconstructing the elided element may lead to a different interpretation (Höhle Reference Höhle, Haider and Netter1991, Bryant Reference Bryant2006: 14–18), which is particularly evident in sentences with quantified NPs as subjects, as discussed in Section 2.2. While (37-a) is interpreted as that the same individual performed two actions, (37-b) is understood as involving two different individuals.

If the analysis with large conjuncts were to maintain, another option is to claim that a non-overt pronoun, pro or PRO, is situated in the SpecCP position of the second clausal conjunct. Neither of both candidates seems to be promising. The first candidate, pro, is unusual because it is not licensed in the Modern Standard German (Jaeggli & Safir, Reference Jaeggli, Safir, Jaeggli and Safir1989). The other candidate, PRO, which is suggested by Munn (Reference Munn1993: 108–110), should be ruled out as well, as assuming a PRO in the subject position of a finite clause does not fall under its distribution. A more general problem about a non-overt pronoun as the subject of the second conjunct is that it is in some cases difficult for this assumed pronominal element to refer to the subject of the first conjunct. In (38), the pronoun in the second conjunct can not receive an anaphoric reading which refers back to the quantified subject of the first conjunct. This applies to coordinate structure in English as well.

Therefore, the above (35) can be better explained under the assumption of phrasal coordination, where two small conjuncts are conjoined and share the same subject located outside the coordinate structure.

The goal of this section is to illustrate that there is no fundamental difference between SLF coordination and phrasal coordination since they exhibit similar structural characteristics and are interpreted in the same way.

4.1. Syntactic structure

First, in an SLF coordination, the prefield of the second conjunct cannot be occupied, i.e. no constituent in the second conjunct can be placed before the finite verb. This applies also to phrasal coordination with a shared subject. In the following example (39), it is not possible to insert a further constituent between the coordinator and the finite verb of the second conjunct.

The second structural characteristic of SLF coordination is that the omission of constituents other than the subject in the second conjunct is not possible. The following examples show that this is equally not a good idea for phrasal coordination. The deletion of other constituents is not possible even when there is no subject gap in the second conjunct.

Next, example (9) shows that in SLF coordination, the adverbial constituents embedded in the first conjunct can take the scope of the whole coordinate structure but cannot exclusively apply only to the second conjunct. The same works for phrasal coordination as well: In (41), the adverb nie ‘never’ is embedded deeply in the first conjunct, but the second phrasal conjunct still remains under its scope.

Contrary to Weisser’s (Reference Weisser2019) view, Höhle (Reference Höhle, Müller, Reis and Richter1983) suggested that both in phrasal coordination and SLF coordination, the scope of adverbs can extend to the second conjunct, as the following sentence (42) exemplifies, so this characteristic was not considered to be unique to SLF coordination.

Finally, it seems that SLF coordination allows asymmetric extraction from the first conjunct. For example, the fronted constituent wen ‘whom’ in (43) below can be regarded as the result of asymmetric extraction from the first conjunct.

Similarly, such unilateral extraction from the initial conjunct is also allowed for phrasal coordination (44).

Empirically, extraction from the first conjunct out of a coordinate structure is not rare and can be attested in other languages as well. It is possible to extract elements from the first conjunct, if there are certain semantic relations between the events described in the conjuncts (Altshuler & Truswell, Reference Altshuler and Truswell2022; Culicover & Jackendoff, Reference Culicover and Jackendoff1997; Kehler, Reference Kehler2002; Na & Huck, Reference Na and Huck1992; Zhang, Reference Zhang2009). In (45), for example, there is one constituent extracted from the first conjunct, suggesting that extraction from the first conjunct is not confined to SLF coordination, either. The German example, (46), is constructed based on (45) and confirms that a similar asymmetric extraction is possible in German.

4.2. Fused interpretation in coordination

In terms of interpretation, it is considered to be a feature of SLF coordination that it is always subject to a fused interpretation (Höhle, Reference Höhle, Müller, Reis and Richter1983; Reich, Reference Reich, Fabricius-Hansen and Ramm2008, Reference Reich2009, Reference Reich, Meibauer, Steinbach and Altmann2013). However, when comparing with the normal phrasal coordination, we can observe that a fused interpretation is not restricted to SLF coordination. In reference to the examples in Section 2.2, we can construct the following sentences which do not involve any subject gaps.

Just as in the previous pair of examples (11) in Section 2.2, the above contrast in (47) shows a similar result. The sentence (47-a) with two conjuncts that share the same subject is not an instance of SLF coordination, but it still receives a temporally sequential and causally related interpretation. It is natural to assume that the two events described in it are related. Sentence (47-b), in which both conjuncts have an overt subject, is rather interpreted as describing two isolated incidents. The interpretation of SLF coordination is just like that of phrasal coordination with a shared subject since they are both coordinate structures that only have one overt subject in the sentence.

A further example shows that the fused interpretation in SLF coordination is not a strict constraint that forces a temporally successive reading. In the following example, the temporal adverbs morgen ‘tomorrow’ and heute ‘today’ specify the time of the events explicitly, so the event in the second conjunct precedes that in the first conjunct.

The temporal sequence of the two events is not a forward one in that the temporally earlier event is in the non-initial conjunct. In regard to such cases, there is no specific restriction on the chronological order of the two conjuncts in SLF coordination. A more accurate description of the fused interpretation is that the conjuncts should be somehow related to each other (Reich, Reference Reich2009). This requirement of relevance is not limited to a single construction as well, but is rather a universal tendency in language use. According to Weisser (Reference Weisser2019), the above example (15) in Section 2.2 shows that it is ungrammatical to conjoin two unrelated predicates within SLF coordination. However, this issue is rather of a pragmatic nature than a syntactic one. On the one hand, as mentioned in Section 2.1, the fronted element in Deutschland ‘in Germany’ takes the scope of the whole coordinate structure, including the second conjunct which states a fact about the freezing point. The sentence may invite a strange interpretation that only in Germany beer has a freezing point of $ -3 $ °C. On the other hand, even if in normal phrasal coordination, conjoining two unrelated predicates is not preferable. Example (49) does not seem to be proper either, even if there is no subject gap in the second conjunct.

One possible alternative explanation is that the subject in sentence (49) has two different thematic roles in the conjoined clauses, leading to the ungrammaticality of the sentence.Footnote ⁴ The following example shows that it is possible for the subject to have different thematic roles in coordinate structures:

Similarly, although the subject in sentence (49) may also involve multiple thematic roles, this in itself does not make it ungrammatical, so the reason for its inappropriateness lies at the pragmatic level. Interpretation of coordinate structures generally involves the establishment of an integrated sentence meaning on the basis of the two conjuncts (Lang, Reference Lang, von Stechow and Wunderlich1991). This general restriction on discourse coherence can be captured by a pragmatic maxim within the classical Gricean framework, the Maxim of Relevance (Grice, Reference Grice, Cole and Morgan1975), which requires conversational contributions to be relevant. Conjoining unrelated propositions in a coordinate structure is generally considered as pragmatically inappropriate, as the (51) show.

Even in a sequence of sentences without any marker of coordination, flouting the Maxim of Relevance is not preferable:

As the problem with the unrelated predicates is of a pragmatic nature, it is possible to construct a context in which such sentences become acceptable. The following sentence, based on (49), can be felicitous:

To conclude, the fused interpretation is not restricted to SLF coordination only. In a narrow sense, the concept of fused interpretation means that the two conjoined events should be considered as an integrated event with a close temporal or causal relationship. In a broader sense, it means that the two clauses should be semantically related in a certain way. However, neither applies exclusively to SLF coordination. Instead, these features hold equally for a coordinated sentence in which the two conjuncts share the same subject.

5. SLF Coordination as a Derivative of Phrasal Coordination

In the comparison above, it has been shown that SLF coordination and normal phrasal coordination share similar syntactic properties and are interpreted in the same manner. In this section, I turn to the differentiating feature of SLF coordination, namely its information structure.

What separates SLF coordination from other instances of phrasal coordination is that the subject of the first conjunct is deeply embedded, which means that the initial position is not taken by the subject, but by another constituent. However, as in the above example (30), an object gap in the second conjunct is not possible in a similar context where a subject gap can be licensed. This contrast hints at possible structural differences between a subject and an object in V2 sentences. There are previous proposals with the assumption that subject-initial V2 sentences in Germanic languages are derived at a lower level of IP, while object-initial V2 sentences are CPs (e.g. Travis, Reference Travis1984; te Velde, Reference te Velde2006; Zwart, Reference Zwart1991). However, these accounts face many empirical problems, see discussions in Schwartz & Vikner (Reference Schwartz, Vikner, Belletti and Rizzi1996) and van Craenenbroeck & Haegeman (Reference van Craenenbroeck and Haegeman2007), and a purely structurally based distinction ignores pragmatic factors, including the role of information structure in determining the constituent in the prefield. One possible solution, as suggested by Mikkelsen (Reference Mikkelsen2015), is to assume that information-structurally differentiated V2 clauses are CPs, while undifferentiated ones are TPs. Although information structure is taken into consideration, this approach still encounters similar empirical problems just as other accounts assuming that the finite verb can remain in the TP. For SLF coordination in particular, example (28) shows that the two conjuncts cannot be below the C’ level (Mayr & Schmitt, Reference Mayr, Schmitt, Everaert and van Riemsdijk2017). I therefore maintain the more commonly accepted position that all German V2 sentences are generated at the CP level, i.e. the finite verb always leaves the TP domain (den Besten, Reference den Besten and Abraham1983; Haider, Reference Haider, Haider and Prinzhorn1986), and integrate information structure into the current proposal by applying the split-CP hypothesis. The aim of the present paper is to establish a direct link between the syntactic structure and the information structure using already existing theoretical assumptions.

Several previous studies (e.g. DeClercq & Haegeman, Reference De Clercq and Haegeman2023; Haegeman & Greco, Reference Haegeman and Greco2018; Mohr, Reference Mohr2009) maintain the assumption that all V2 sentences are derived at the CP level. By adopting a cartographic approach based on Rizzi (Reference Rizzi and Haegeman1997), these authors provide a more fine-grained structure for the left periphery of V2 sentences in Germanic languages. My analysis proceeds on the basic assumption in Mikkelsen (Reference Mikkelsen2015) that information structure affects the structure of a V2 sentence, and combines it with a refined structure of the left periphery. Following Meelen et al. (Reference Meelen, Mourigh, Cheng, Bárány, Biberauer, Douglas and Vikner2020) and Walkden (Reference Walkden2017), I propose that the CP in German can be split into two layers: a lower FinP and a higher Foc/TopP which conflates CP layers with a special pragmatic effect, including FocP and TopP.Footnote ⁵

Within this structure, the sentence-initial constituent is in the SpecFinP position in V2 sentences with an undifferentiated information structure, and in information-structurally differentiated V2 sentences, the sentence-initial constituent is moved to the higher SpecFoc/TopP position.

In terms of the structure of coordination, I assume that all coordinate structures have the identical syntactic representation as proposed by Munn (Reference Munn1993), which can be schematically illustrated as follows in (55). This structure features asymmetry, in which the second conjunct is considered as an adjunct to the first conjunct and thus constitutes an adjunct island. Munn (Reference Munn1993) mainly discussed cases where maximal projections are coordinated, as (55) shows, and intermediate projections are neither discussed nor made available by the machinery of this model, which was later criticised by Borsley (Reference Borsley2005). The present analysis maintains the core assumption of Munn’s model of coordination that the second conjunct is an adjunct to the first conjunct, but follows a later line of work in allowing coordination to target intermediate projections, which will be discussed below.

Since the second conjunct is an island, extraction out of it should thus be disallowed. One notorious exception, as already noted in Ross (Reference Ross1967), is the across-the-board-movement (ATB-movement), where a single element is considered to have moved out of multiple positions in a coordinate structure:

There are already many theories attempting to explain the ATB-movement.Footnote ⁶ For example, Munn (Reference Munn1992) proposed that ATB-movement shares the same mechanism as parasitic gaps. German, however, does not have parasitic gaps, so Reich (Reference Reich2007) proposed that ATB-movement is the result of the asymmetric extraction out of the first conjunct and the simultaneous deletion of its copy in the second conjunct. For present purposes, we will set aside the details of the discussion on the mechanism of ATB-movement itself, as it suffices for us to be able to assume that the movement of the subject can also be across-the-board. Therefore, subject movement out of both conjuncts into the matrix FinP is allowed, even though independent extraction from the second conjunct is not. For phrasal coordination, the subject in the prefield is situated outside the coordinate structure and it is shared by both conjuncts as the result of an ATB-movement of the subjects. Burton & Grimshaw (Reference Burton and Grimshaw1992) and McNally (Reference McNally1992) proposed the following derivation of phrasal coordination with a shared sentence-initial subject, as shown in (58), an instance of VP coordination. Drawing on the VP-internal subject hypothesis (cf. Kitagawa, Reference Kitagawa1986; Koopman & Sportiche, Reference Koopman and Sportiche1991), they followed the assumption that all subjects are generated within the VP, where the assignment of theta-roles takes place (McNally, Reference McNally1992). In the next step, the subjects go through ATB-movement to their surface position in the prefield.

This solution can be extended to the coordination of larger constituents, like I’-level coordination (Burton & Grimshaw, Reference Burton and Grimshaw1992; McNally, Reference McNally1992):

In the current analysis of SLF coordination, I propose that the conjuncts are Fin’ rather than FinP. By conjoining two Fin’ nodes, the subject in SpecFinP can successfully c-command both conjuncts, satisfying the requirements for subject sharing in a split-CP framework, given that the parasitic gap analysis of ATB-movement in Munn (Reference Munn1992) is not adopted here. The assumption that intermediate projections like V’s, I’s or Fin’s are conjoined, although a common one (e.g., Heycock & Kroch, Reference Heycock and Kroch1994; Höhle, Reference Höhle, Mascaró and Nespor1990; Johnson, Reference Johnson2002; Mayr & Schmitt, Reference Mayr, Schmitt, Everaert and van Riemsdijk2017), faces a potential theoretical issue that it collides with the assumption that a maximal projection only has one head.Footnote ⁷ One way to reconcile the two theoretical assumptions is to further assume that there are multiple intermediate levels between CP and TP. For example, Borsley (Reference Borsley2005) suggested that by adopting the split-CP hypothesis (Rizzi, Reference Rizzi and Haegeman1997), the conjuncts in sentences like (59) and (60) could be maximal projections of different C categories. Bošković (Reference Bošković, van Craenenbroeck, Pots and Temmerman2020) also analysed examples like (61) as TP coordination with the subject moved to a higher level:

Another option is to give up the theoretical dichotomy between maximal and intermediate projections, but to divide elements into projecting and non-projecting ones, as suggested by Zhang (Reference Zhang2006). Within this framework, coordination does not target maximal projections only, but can also apply to lower levels as well. This approach not only aligns with the plausibility of treating Fin’ as the domain for coordinate structures after ATB-movement in the present analysis, its flexibility also allows lower-level coordination, such as head coordination and prefix coordination, to be captured in the same model. In (62), even though the second conjunct a Tibetan folk song is larger than a word or can even be treated as a DP, Zhang (Reference Zhang2006) argued that conjuncts are non-projecting elements since the projectability is determined relationally via Merge, not intrinsically. Under this view, intermediate projections including V’, I’, Fin’ are in fact non-projecting complements to the coordinator. The resulting coordinated structure is a maximal projection headed by the coordinator, not by either conjunct’s internal head.

Similar to (59) and (60), we can analyse phrasal coordination in German as follows, where the subject underwent ATB-movement and left the coordinate structure:

SLF coordination is derived on the basis of the above structure. Fronting a non-subject element to the prefield involves contrastive topicalisation or focalisation, which is steered by information structure and the corresponding C head contains a strong fronting feature that needs to be filled (Gärtner & Steinbach, Reference Gärtner and Steinbach2003). The topicalised or focalised constituent is extracted asymmetrically from the first conjunct, and the landing site of the moved element is SpecFoc/TopP. An extraction from the second conjunct is not possible due to its islandhood.Footnote ⁸

This structure explains the traits of SLF coordination and phrasal coordination. First, in SLF coordination, the prefield of the second conjunct must remain unoccupied, because the second conjunct is a Fin’ and the SpecFinP does not fall under the scope of the second conjunct. Second, the omission of the object and other VP-internal constituents is not licensed in the second conjunct since the corresponding elements are not fronted and remain embedded in the first conjunct. This is not the case for the subject, which is out of the coordinate structure. Third, adverbial constituents in the first conjunct or at the fronted position are structurally higher and therefore can take the scope of the whole sentence. Finally, extraction is only possible from the first conjunct, because the second conjunct is an island. These deductions are in line with the previous theories that assume an asymmetric structure for coordination or analyse the second conjunct of an SLF construction as an adjunct.

In terms of the semantic relation between the conjuncts, we can conclude that there is no difference between an instance of SLF coordination, which is a Foc/TopP, and its corresponding coordinated sentences without a subject gap, which are derived at the FinP level. The distinctive point between them lies in the information structure, i.e. a non-subject element is contrastively focalised in SLF coordination. The fused interpretation is not limited to SLF coordination. Instead, it is also featured by normal phrasal coordination with a shared subject in the prefield. But there seems to be a counterexample to this:

According to Büring & Hartmann (Reference Büring and Hartmann1998), the sentence above is an instance of phrasal coordination and has two possible interpretations. The sentence can be interpreted as describing two different individuals as in (65-a), or as describing two different traits of the same individual (65-b). Based on our world knowledge, we would consider sentence (65-a) to be more likely true, but (65-b) is probably false because it is unlikely that the same individual could hold both positions at the same time. According to Büring & Hartmann (Reference Büring and Hartmann1998), the two propositions differ in terms of their truth-conditional meaning, but both readings are accessible. In contrast, the SLF coordinated sentence (66) only has the fused interpretation as in (65-b).

However, as the previous examples in (37) already show, it is not a usual case that phrasal coordination can have both a fused and a disassociated interpretation. Sentence (67-a), which is adapted from (35), only has a fused interpretation, where the same woman performed two actions consecutively, and it clearly contrasts with sentence (67-b), where two different individuals are involved.

This seeming counterexample (65), in fact, supports the current proposal based on information structure, as its two interpretations are closely related to the focalisation of the fronted element. The two different interpretations can be realised by putting the focalised stress on different elements. The reading involving two different agents is triggered by a specific information structure and does not directly map to the truth-conditional meaning of the sentence. For the interpretation (66-a), the subject in the prefield should receive a contrastive stress as in (68):

Following Rooth’s (Reference Rooth1992) alternative semantics, placing focus on the DP means that a set of alternatives is contextually possible (Wagner, Reference Wagner2012). Specifically for (68), the possible alternatives include individuals which are not female. This focused sentence-initial subject does not denote a specific female individual, but rather the kind. Its actual interpretation is similar to (69), a statement about female politicians in general, where a bare plural is used to give it a generic reading, as suggested by Cohen & Erteschik-Shir (Reference Cohen and Erteschik-Shir2002):

Similarly, the focused subject in (68) is generic and it refers to a kind, instead of a specific individual. On the basis of our world knowledge that the two positions will not be held by the same person, the sentence encourages an implicature that two different women are being referred to, although whether the two referred individuals are identical is not part of the literal meaning of the sentence at all. The underlying intention of this utterance is normative and metalinguistic, which is to propose a change in the underlying beliefs about a specific group people as a whole (Hesni, Reference Hesni2024), as in the following example, where the subject is also used generically as the predicate real highlights a potential contrast with those who are not real:

By contrast, if the subject is not focused, it stays at the SpecFinP position, and the fused interpretation, as in (65-b), is the preferred interpretation for (65). Following Mohr (Reference Mohr2009) and Mikkelsen (Reference Mikkelsen2015), I argue that the sentence-initial subject with a contrastive focus is moved to the higher CP layer, SpecFoc/TopP. If the fronted element does not receive a contrastive intonation, it remains in the SpecFinP position and the sentence has an undifferentiated information structure. This is exactly the case in (66), where the subject stays in the lower SpecFinP position, which is in line with the observation in Cohen (Reference Cohen2001) that singular indefinites are not generic when they are not topicalised.