Complementizer Agreement and the Licensing of DPs: An Account in Terms of Referential Anchoring

Abstract

In this paper, I argue that the phenomenon of complementizer agreement in West Germanic and the distribution of DPs in German can be given a common explanation in terms of an approach in which context values are not freely assigned via an interpretive function operation, as is assumed in standard accounts of formal semantics, but rather, they become accessible in a specific functional head in the C-domain.

1. Introduction

In this paper, I address the well-known issue that DP-arguments, dependent on their interpretation, occupy different domains in the clause in many languages. For instance, definite DPs in German occupy higher positions in the middle field than indefinite ones (cf. Kratzer 1989; Diesing 1992). It is assumed that indefinite DPs can be licensed in the V-domain, while definite ones move higher and are presumably licensed in the T-domain.1

In addition, discourse anaphoric DPs and pronominal DPs move to even higher positions in the clause in German and can be argued to be licensed in the C-domain. Furthermore, there is the phenomenon of complementizer agreement in many Germanic dialects that is still lacking an intrinsic motivation. It is generally assumed that a C-head enters in an agree relation with a finite verb (in T). However, this remains a stipulation. I will argue in this paper that complementizer agreement is a reflex of an intrinsic licensing relation between Fin⁰ and the temporal argument structure of the verb, on the one hand, and referential DPs, crucially including subject pronouns, on the other hand.

We may ask what the reason for the movement of definite DPs (and pronouns) is, given that the standard semantic account of the interpretation of referential expressions is in terms of assignment functions that assign a referential index to a DP from a context set. If this were correct, DP-licensing would be possible in any position of the clause.

I note that case or agreement with a finite verb cannot be taken to motivate these movement operations either, since indefinite DPs also have case and can, as subjects in a presumed vP internal position, enter into an agree relation with T (or AgrS, depending on one’s favorite theoretical assumptions).

I will argue that the reason behind this distribution is that context values are not freely assigned but become accessible only in the C-domain that, as advocated by Rizzi (1997), serves to connect the proposition with the context.

In particular, I will assume that context values, that is, values for established discourse referents, on the one hand, and for the reference situation that is crucial for the temporal anchoring of the clause, on the other hand, are accessible in Fin⁰.

Furthermore, I will argue that the more fine-grained distribution of indefinite DPs (in the V-domain), weak (definite) DPs (in the T-domain), and strong (definite) DPs (in the C-domain) follows from the presence or lack of presuppositional requirements that the different determiners impose on the individual and the situation argument of the nominal head.

2. DP Types and Their Role in the Discourse

Let us start out with a discussion of the different discourse roles of definite and indefinite DPs. It was proposed first by Irene Heim (1982) in her famous familiarity condition that indefinite DPs serve to introduce new discourse referents, while definite DPs serve to pick up (i.e., refer back to) referents that have been introduced in the previous discourse. While this is certainly correct for typical cases of the use of definite DPs, it was soon after noticed that there are uses of definite DPs that also introduce a new discourse referent.

In fact, Donnellan (1966) had pointed out that one needs to distinguish between the referential use of a definite description, which obeys Heim’s familiarity condition, and the attributive use of a definite description, in which it is essential that there be a uniquely identifiable referent (typically not yet given in the existing discourse) in the situation that is at issue, as in (1). The referent of the winner in (1) is not known at the point of the utterance, but it is uniquely identifiable as soon as the race has been decided.

1.	Tomorrow there will be a 100 m run in Vienna.
	The winner will receive a Porsche!

Thus, there are two different conditions, the familiarity condition and the uniqueness condition, that both seem to be relevant for the use of the definite determiner. The dispute, in fact, goes back to Russell (1905) and Frege (1892). While the former proposed that the definite determiner purports an assertion as to the existence of a unique individual that fulfills the nominal predicate and is, thus, apt to account for the attributive use of a definite description, Frege (1892) held that a definite description imposes a presupposition that there exists an individual that fulfills the nominal predicate.

As far as (1) is concerned, Russell’s treatment would foresee the complex assertion in (2), while Frege would argue that the presupposition of the definite description in (1) is fulfilled, since when there is a race there is a unique winner, and (1) amounts to the assertion that this individual (the one that fulfills the presupposition) will receive a Porsche.

2.	ιx winner (x) & will-receive-a-Porsche (x)

I will assume in this paper that both are right, and I will distinguish between the weak and the strong definite determiner in Germanic. It has long been noted that several Germanic languages/dialects have two full article paradigms (cf. Heinrichs (1954) for Rhineland dialects, Scheutz (1988) and Schwager (2007) for Bavarian, and Ebert (1971) for the Frisian dialect of Fering). In standard German, the distinction becomes apparent in certain preposition–article combinations, as is illustrated in (3):

3.	a.	Hans	ging	in-s	Haus.	(D-weak)
		John	went	into the	house.
	b.	Hans	ging	in das	Haus.	(D-strong)
		John	went	into the	house.

While (3a) can be uttered out of the blue, (3b) is only possible if the relevant house has already been mentioned in the previous discourse. Thus, it appears that we have to deal with two different types of definite determiners that also differ in their semantics. Schwarz (2012) argued that one should not strive for a unified theory of the semantics of the definite determiner in Germanic since the weak definite determiner is subject to a uniqueness requirement, while such a requirement is apparently irrelevant for the strong definite determiner. Readers are referred to Schwarz (2012) for further details.

Following Frege (1892), I propose that the uniqueness condition is also relevant for the strong determiner but only in the definition of the presupposition that serves to discriminate the antecedent of the strong definite DP in the discourse.

Furthermore, I will make the following proposal to solve the question about the definition of the situation in which the uniqueness condition imposed by a definite determiner is supposed to hold. While in standard treatments of definite descriptions, as in Schwarz (2012) and others, it has been assumed that a situation pronoun is introduced by a definite determiner, and hence, it is absent in indefinite DPs, I propose that this situation argument is introduced already by the nominal head, and hence, it is also available in indefinite DPs.

In other words, every nominal referent is individuated with respect to a situation. However, definite DPs, both weak and strong ones, impose a presupposition on the identifiability of this situation argument in the common ground (CG), while indefinite DPs come without any presupposition on this argument. This means that a definite determiner indicates that the situation argument of a nominal is, in some sense, given, while an indefinite determiner indicates via an implicature that the situation argument of a nominal is not given in this sense. Thus, both weak and strong definite determiners share the property of imposing the same usage condition on a definite description that is distinct from the usage condition of an indefinite description.

The difference between weak and strong definite determiners only concerns the individual argument of a nominal. With a weak determiner, it is asserted that there is a unique individual in a situation given in the CG, while with a strong determiner, it is presupposed that there is a unique individual identifiable in the CG, which satisfies the nominal predicate in the situation given in the CG, as illustrated in (4). In (4), conditions that operate as presuppositions are put before the dot of the lambda-operator and are underlined, while conditions that are asserted appear after the dot of the lambda-operator.

4.	a.	⟦D⟧ = λP ∃s s in CG . ιx P (x, s)	(weak definite determiner)
	b.	⟦D⟧ = λP ∃s s in CG & ιx in CG & P (x, s). x	(strong definite determiner)

This means that if an indefinite DP is merged in the vP, its situation argument, being without any presupposition, can be identified with the event argument of the verb, while a definite DP cannot be licensed in the vP since its presupposition requires that its situation argument is identified with a situation that is already in the CG. One specific situation that is already in the CG and becomes available in the T-domain is the reference situation that plays a crucial role in the temporal interpretation of the predicate, as we will see in the next section.

As far as pronouns are concerned, I propose that they also have a nominal core, namely, the abstract nominal predicate participant that relates an individual argument and a situation argument, as is illustrated for a personal pronoun in (5) and for a deictic pronoun in (6). The features in D are interpreted as presuppositional conditions on an individual referent that serve to discriminate the discourse antecedent in the context. The situation argument of personal pronouns is identified with the reference situation (s_R) (see below), while the situation argument of deictic pronouns is identified with the utterance situation (s_U). These feature specifications require pronouns to be licensed in the T-domain for their situation argument and to have access to the C-domain for their individual argument.

5.	he = [DP D [nP participant (x, sR) ]]
	3SG, male
6.	we = [DP D [nP participant (x, sU) ]]
	1PL

Summing up this section, I note that Diesing’s account can explain why indefinite DPs need to be licensed in the V-domain, but it fails to account for why definite DPs must move out of the V-domain. The present account, on the other hand, explains why indefinite DPs can be licensed in the V-domain and why definite DPs must move out of the V-domain to be licensed in the T- or C-domain. In particular, I will argue that weak definites are licensed in the T-domain while strong (referential) definites are licensed in the C-domain.

3. The Reference Situation and the Anchoring of a Thetic Judgment

It is standardly assumed that a predicate is anchored via tense (and mood) to a context. Interested readers are referred to Zagona (2013) and the references cited therein for further background on this issue. In particular, it is assumed that tense (in a matrix clause) locates the verbal event with respect to the utterance time. Thus in (7), a speaker asserts that there was an event of visiting in the past (before the utterance time) in which an individual named John functioned as the agent (the visitor) and his mother functioned as the theme (the visitee of the event).

7.	a.	John visited his mother.
	b.	∃e visiting(e) & past (e) & agent (e, John) & theme (e, his mother)

However, this simple linking approach is insufficient when we look at examples embedded in a discourse, as illustrated in (8). Anaphorically linking she to his mother in (8a), and simply anchoring the event e₂ to the utterance time, the meaning of (8b) would be compatible with this event preceding, following, or overlapping with e₁ as long as both events precede the utterance time (s), as illustrated in (8c). This rendition is incomplete, since speakers typically interpret (8b) as a claim about John’s mother being sick at the time of his visit. Moreover, the adverbials in (8d) are also interpreted with respect to John’s visit.

8.	a.	John visited his mother. (e1)
	b.	She was sick. (e2)
	c.	e1 < e2 < s, e2 < e1 < s, or e1 o e2 < s
	d.	She was sick one week before/earlier.

The problem can be solved by introducing a reference situation. Here, I am following Reichenbach (1947), according to whom tense establishes a link between a speech time and a reference time, as illustrated in (9). The event argument of the verb is then taken to be situated with respect to the reference time by aspect, as is illustrated in (10).

9.	The meaning of tense according to Reichenbach (1947):
	a.	past: = r < s
	b.	present: = r ⊆ s
10.	The meaning of aspect according to Reichenbach (1947):
	a.	perfect: = e < r
	b.	imperfect: = e ⊂ r

As is indicated in (8), the event in (8a) serves as discourse antecedent for the interpretation of tense in the clause in (8b) and for the temporal adverbials in (8d). Hence, I will propose that tense is not a predicate of a point of time or an interval, as is standard since the seminal work on tense by Stowell (1995), but rather, that tense relates two situations (an utterance situation and a reference situation). Hence the temporal interpretation itself is secondary and derived from a relation between situations making use of the running time of a situation (τ), as illustrated in (11).

11.	Situation-based account of tense (SAT):
	-	Tense is a predicate that relates situation arguments.
	-	past (s1, s2) = s1 precedes s2 = : τ (s1) < τ (s2)

Normally, a predicate is anchored to a context by a definite subject, as in (12a). The result is a categorical judgment about a particular individual (or a particular set of individuals). Alternatively, a predicate can be anchored to a context via a reference situation, as in (12b). The result is a thetic judgment about a particular situation. It was argued in Hinterhölzl (2024) that es in German is not an expletive element but (being inserted in [Spec, TP]) instead binds a reference situation argument of tense. An anonymous reviewer asked why es in (12b) cannot be taken to be inserted in the C-domain since it serves to distinguish between a judgment and a question, properties that are taken to be defined in the C-domain. I assume that sentence types (declarative or interrogative moods) are defined in ForceP and that FinP, together with TP, serves to referentially anchor a proposition.

12.	a.	Hubert Haider spricht.
		Hubert Haider speaks.
	b.	Es spricht Hubert Haider.
		It speaks Hubert Haider.

More importantly, es becomes obligatory if a subject semantically cannot serve as an anchor because it is indefinite as in existential constructions, as in (13a), or if a predicate does not have an argument of itself, as with weather verbs, in which case the verb is predicated of the reference situation argument of tense, as illustrated in (13b–d). In (13d), s_U refers to the utterance situation. Thus, both (13a) and (13b) constitute thetic judgments about a specific situation.

13.	a.	Es gab einen Aufruhr.
		It gave an uprise (there was an uprise).
	b.	Es     regnete.
		It        rained.
	c.	s1 (that is identified with the reference situation) ∈ ⎨s | rains in s⎬
	d.	λP(s,sU). ιs P(s,sU)                             (meaning of es)

In English sentences with an indefinite subject, the adverbial there is inserted in Spec, TP. Also, here, I argue that there is not an expletive element but instead serves semantically as an alternative anchor in the clause, as illustrated in (14).

14.	a.	John visited his mother.
	b.	There was a child crying in the garden.
	c.	I went to the local bar last night. Into the room walked a man with a green hat

In the present account, there is a function that maps the reference situation onto its location and refers back to the situation of John’s visit that provides the situation with respect to which the predicate ‘was a child crying in the garden’ is temporally and locally evaluated. In a similar vein, a PP, by denoting the resultant location of a predicate expressing a change in state (or location), can serve as a subject/anchor in the case of locative inversion, as illustrated in (14c). In (14c), into the room refers to the room in the previously mentioned bar situation from the previous night. Thus, both (14b) and (14c) qualify as thetic judgments.

As already argued for by Milsark (1974), there cannot be treated as an expletive element that is replaced at LF by a real subject, and this is shown by the observation that the subjects in (15) have the following different interpretations: in (15a), the subject has a weak cardinal interpretation, and in (15b), the subject has a strong proportional interpretation. Interested readers are referred to Hinterhölzl (2019) and the references given therein for a thorough account of the syntax and semantics of the expletive construction in English.

15.	a.	There were not many people in the room.
	b.	Many people were not in the room.

It is interesting to note that strong quantifiers can anchor a predicate to a context but need not do so, as illustrated in (16). In (16), taken from Schwarz (2012) and also discussed in Hinterhölzl (2019), the subject most senators has a strong proportional reading, but the sentence appears to characterize the political situation in 2004, constituting a thetic judgment.

16.	What was the political situation in congress in 2004?
	Most senators were Republicans.

Arguably, we have a case in which an attributively used DP is evaluated with respect to a given situation, hence the strong interpretation. I propose that the subject in this case is licensed in the T-domain by identifying its situation argument with the reference situation of tense. The speaker in (16) does not make reference to a specific group of senators, but rather, is simply stating that a majority of the senators at that time were Republicans.

In conclusion, we propose that indefinite DPs and weak quantifiers are interpreted in the V-domain while attributively used definite DPs and strong quantifiers (if not discourse-anaphoric) are interpreted in the T-domain. Furthermore, I propose that referentially used definite DPs and anaphoric strong quantifiers require access to the C-domain to be fully licensed. In particular, I propose that discourse-anaphoric DPs must enter into a licensing relation with the head Fin⁰.

In the following section, I will discuss the distribution and the licensing of subjects in Cimbrian. These data will provide an interesting parallel to the patterns found in complementizer agreement in languages/dialects that allow for double agreement.

4. Subjects in Cimbrian: A Case Study

Let us take a look at the distribution of subjects in Cimbrian, a German dialect spoken in the village of Luserna, Trentino. Field work was carried out by Federica Cognola, and the data were published in Cognola and Hinterhölzl (2020). Interested readers are referred to this article for a more complete picture of the complex interaction between V2, question formation, and the licensing of subjects in this variety. As is illustrated in (17), there is a complementary distribution between preverbal subjects and the presence of a subject pronoun, or da (‘here, there’), cliticized on to the verb. In (17), da is spelled out as -ta when cliticized on the verb. The subject in (17f) is unmarked with respect to its information structural role (i.e., it can be new or given information), as follows:

17.	a.	Bas hatt-arj herta gekoaft dar Lucaj?	d	*Dar Luca hatt-arj herta gekhoaft in libar.
		What has he always bought the Luca?		The Luca has he always bought a book.
	b.	Bas hat-ta herta gekoaft dar Luca?	e.	*Dar Luca hat-ta herta gekhoaft in libar.
		What has_da always bought Luca?		The Luca has da always bought a book.
	c.	*Bas hat herta gekoaft dar Luca?	f.	Dar Luca hat herta gekhoaft in libar.
		What has always bought the Luca?		The Luca has always bought a book.
		“What has always Luca bought?”		“Luca has always bought a book.”

This characterization can also be found in various publications about the role of da in Cimbrian (Bidese and Tomaselli 2005 and subsequent work; Kolmer 2005; Grewendorf and Poletto 2015), where da and a subject clitic are ruled out in all cases in which a subject precedes a finite verb, and da or a subject clitic are obligatory in all cases in which a subject follows a finite verb.

As I will argue below, movement of a constituent into the preverbal domain interferes with the licensing of subjects. This is due to the V2 nature of the language where [Spec, FinP] constitutes a bottleneck for movement into the C-domain.2 In other words, the wh phrase bas in (17) has to pass through [Spec, FinP] to reach [Spec,ForceP] to license the speech act of a question. Subjects in this case can be licensed in a lower position when they are doubled by da or by a subject clitic pronoun, as follows: if doubled by da, the subject is focused (new information or a contrastive focus), and if doubled by a pronoun, the subject is a topic. The main stress in (18a) falls on the sentence final subject while the sentence final subject in (18b) is optional and unstressed.3

18.	a.	Haüt iz = ta khent dar nono
		Today is da arrived the grandfather.
	b.	Haüt izz = arj khent (dar nonoj)
		today is = he arrived (the grandfather)
	c.	*Haüt iz khent dar nono
		Today is arrived the grandfather.
		“The grandfather arrived today.”

We may wonder what the roles of clitics and da are in the licensing process of the lower subject in (18ab). Let us first discuss what is said about da in the literature. Da is only homophonous with the locative da (‘here’) (see Grewendorf and Poletto 2015, p. 402; Kolmer (2005); and Bidese and Tomaselli (2018 and previous work)). As illustrated in (19), da cliticized onto a main verb can occur with an instance of the locative da.

19.	Bas hat-ta gatont a khin da?	(Grewendorf and Poletto 2015, p. 402)
	What has da done a boy there?
	“What has a boy done there?”

Da differs from the English ‘there’, realizing [Spec,TP], since da is compatible with definite and indefinite NPs and nothing can intervene between da and a finite verb in main clauses and between da and the complementizer bo (cf. Bidese et al. 2012).

Furthermore, it is uncontroversial that da is hosted in the lower portion of CP, i.e., FinP (see Rizzi 1997), and that its position with respect to a finite verb is fed by V-to-C movement (see Bidese and Tomaselli 2005 and subsequent work and Grewendorf and Poletto 2015 for an analysis of Cimbrian as a V2 language).

The idea that I would like to develop in the following section is that clitics and da, by undergoing head movement to Fin⁰, serve to license the subject when the latter is unable to undergo movement to [Spec, FinP] for syntactic (i.e., another element moves through [Spec, FinP]) or semantic reasons (i.e., the subject is indefinite).

5. The Special Role of Clitic Pronouns and da in Anchoring the Utterance to the Context

In this section, I will argue that clitics and da serve to referentially anchor a subject to a context. In particular, I propose that definite DPs, in contradistinction to indefinite DPs, have an extra layer, which is an additional functional head that licenses a correlate DP in its specifier, as in (20).

Indefinite and weak DPs lack the respective layer and, thus, must combine with an adverbial alternative anchor, as occurs with the English existential ‘there’ construction.

20.	[DP [DP da/cl] [D0 [NP N]]

Referential subjects can always anchor a predicate and obtain access to the value of their discourse antecedent if they move into a pre-finite position, that is, into [Spec, FinP] in Cimbrian, as illustrated in (21). Furthermore, I propose that if a referential DP is moved into [Spec, FinP], no correlate is generated in [Spec, DP] for reasons of economy, explaining the ungrammaticality of (21b).

21.	a.	Dar Mario hat gekhoaft in liber.
		The Mario has bought the book.
	b.	*Dar Marioj hat-ta/hat-arj gekhoaft in liber.
		The Mario has da/has he bought the book.
		“Mario bought the book.”

Non-definite subjects can anchor an utterance to a context if the DP has a strong interpretation (QN or WhN), as illustrated in (22). Also, in this case, a correlate clitic or da are excluded for reasons of economy.

22.	a.	Belz khinn hatt bokhennt soin tatta?
		Which child has met his father?
	b.	*Belz khinn hatt-ta bokhennt soin tatta?
		Which child has da met his father?
		“Which kind met his father?”

With non-definite subjects without any NPs overtly realized (e.g., bare QPs and simple wh elements), the following two cases must be distinguished: (a) non-subject questions trigger always the presence of a clitic or da depending on whether the lower subject is to be interpreted as a topic or as a focus, as illustrated with a focused subject again in (23).

23.	Bas hat-ta herta gekoaft dar Luca?
	What has da always bought the Luca?

Here, the idea is that since the subject is blocked by wh movement to move into [Spec, FinP], the correlate sub-extracts from the subject DP, which remains in a lower position and undergoes head movement to Fin⁰ to connect the subject with a specific referential value from the context set, where the features of the clitic serve to discriminate the relevant discourse antecedent.

With subject questions, da is optional depending on the interpretation of the subject, as illustrated in (24) and (25). Since the wh subject is moved through [Spec,Fin], it alone can anchor the utterance if it has a definite interpretation, as in (24). If it has an indefinite interpretation, the adverbial da must be inserted in [Spec,TP] and undergo head movement to Fin⁰.

The data in (24) and (25) are taken from an empirical investigation, i.e., interviews with native speakers of Cimbrian in Luserna carried out by Federica Cognola and reported in Cognola and Hinterhölzl (2020). Interested readers are referred to this paper for the details of this study. In (24) and (25), the native speaker’s judgments are given by an evaluation on the Likert scale between zero (ungrammatical) and five (fully grammatical). Since the context in (24) triggers an indefinite interpretation on the wh subject, only the version with da is possible. Since the context in (25) triggers a definite interpretation on the wh subject (who of us), the version without da is fully grammatical.

24.	Context: You are watching TV and hear the telephone ringing. You ask the following:
	a.	Ber riüft-ta         o?            → 4,8/5
		Who calls da      up?
	b.	Ber riüft    o?             → 2/5
		Who calls           up?
		“Who is calling?”
25.	Context: You and your friends have to book a room for the weekend. You do not
	know who is supposed to call the hotel. You ask the following:
	a.	Ber riüft-ta                o?                → 2,5/5
		Who phone da    up?
	b.	Ber riüft     o?                → 4,8/5
		Who phone         up?
		“Who of us is going to make the call?”

An anonymous reviewer asked why it is that only subjects interact with clitic pronouns and da in Cimbrian, while it must be assumed that all referential DPs (of objects and prepositional objects) must have access to Fin⁰. The latter assumption is correct. I propose that Fin⁰ enters into an agree relation with all the referential constituents contained in TP and values them, but it will only attract the referential subject since it constitutes the highest argument in the structure. Given that it is the movement of the subject into [Spec,FinP] that interferes with the wh movement via the bottleneck effect, it is subjects that interact with the presence/absence of clitic pronouns and da in Cimbrian. When a subject is non-referential (i.e., indefinite), it will remain in a lower position and the sentence will be anchored via the reference situation argument of tense as a thetic judgment. Here, I will leave aside the issue of anchoring referential adverbials.

To sum up what we have found so far, referential DPs cannot be interpreted within vP without any additional operation that connects them with the C-domain (see (17b) above, repeated here as (23)). I have proposed that referential DPs have an extra layer. A definite strong DP, when unable to move to [Spec, FinP] for syntactic reasons, is licensed by movement of a correlate adnominal da or a pronominal correlate, and the choice is language- or function-specific (topic vs. focus). Note in particular that some languages also allow clitics with focused constituents, such as Spanish and Romanian. However, I cannot address this issue in any detail in this paper. In the following section, I will address the question of what happens in cases where the subject is not anchored via an anaphoric link.

6. The Role of Frame Adverbials

Frame adverbials play a special role in the anchoring process of statements. I first note that IP-related temporal (and locative) adverbs express a relation between the reference situation and the event time/location, as we have already seen in (8d) above.

Frame adverbials crucially have a different interpretation. They shift or restrict the reference situation itself, as illustrated in (26) and (27). While speaker A in (26) talks about Christmas in the past, speaker B shifts the reference point with the expression ‘in not many years’ to a future reference situation. Likewise, in (28), the adverbial ‘with no job’ restricts the set of people that would be happy, giving rise to a strong proportional reading of the quantifier ‘few’.

26.	A: Last year, Christmas was fun. We had 5 days of free holidays.
	B: In not many years, Christmas will fall on a Wednesday again.
27.	With no job, few people would be happy.

An initial check of Cimbrian data indicates that generic statements in Cimbrian always appear without da or a subject clitic, but this was not investigated in detail by Cognola and Hinterhölzl (2020). If this observation is verified on a larger dataset, it would imply that the subject is anchored in a different way in these cases.

Here, I will limit myself by motivating this claim with English data. As illustrated in (28), a frame adverbial such as ‘in Australia’ restricts the set of swans to Australian swans. I propose that in this case, the situation argument of the adverbial binds the situation argument of the nominal subject, leading to the interpretation indicated in (28c). Since the subject is interpreted with respect to a new (but anchored) reference situation, definite DPs can only have a weak interpretation. This bleeds the necessity of entering into a relation with Fin⁰ for the assignment of a context value for the individual argument of a subject. Thus, I conclude that subjects in the presence of frame adverbials do not need to be anchored by Fin⁰ since they receive a bound interpretation.

28.	a.	Swans are white.
	b.	In Australia, swans are black.
	c.	Gx in Australia (s1) and swans (x, s1) → black (x)

In the following section, we will see that the pattern of subject licensing in Cimbrian is replicated in systems of complementizer agreements in West Germanic dialects.

7. Complementizer Agreements in West Germanic

As van Koppen (2016) showed, complementizer agreement (CA) is a complex and manifold issue in West Germanic. This is illustrated in (29–31). If a subject is focused, some dialects show CA, such as Austrian Bavarian in (29), and some dialects lack CA, such as Hellendoorn Dutch (30), and there are dialects in which a sentence is ungrammatical with or without CA, such as Frisian (31).

29.	Warum-st grod DU mein Freind net griasst ho-st, vasteh i a net.
	Why-2P.SG PRT you my friend not greeted have-2P.SG. understand I too not
	Why you of all people didn’t greet my friend, I don’t understand either.’
	(Bavarian, Gmunden dialect, Gruber 2008, p. 53)
30.	dat/* darr-e [zölfs wiej] de wedstrijd wint
	that/that-Agr even we the game win      (Hellendoorn Dutch, van Koppen 2012)
31.	a.	* Hy leaude dat-st moarn do komme soest
		he believes that-2P.SG tomorrow you come should-2P.SG
	b.	* Hy leaude dat moarn do komme soest
		he believes that tomorrow you come should-2P.SG
		(Frisian, Germen de Haan p.c., Fuß 2008, p. 85)

van Koppen (2012) argued that there are two types of CA to be distinguished. In a type A dialect, such as the dialect Tegelen Dutch, an agreement suffix is similar to an agreement suffix on a verb and CA is insensitive to subject movement and to subject modification. In a type B dialect, such as Hellendoorn Dutch, an agreement suffix differs from an agreement suffix on a verb, displaying the phenomenon of so-called double agreement (DA). Furthermore, in these dialects, agreement suffix is of pronominal origin and CA is sensitive to subject movement and subject modification. I will illustrate the variable nature of DA that occurs in various West Germanic dialects with van Koppen’s (2012) data from Hellendoorn Dutch.

It is illustrated in (32) that if a subject is moved into a preverbal position, DA (-e) is ruled out, while if a subject stays in a lower position, as in a yes/no question that requires a V1-order, DA is necessary. It is illustrated in (33) that if a subject is modified by a focus particle, DA is ungrammatical, and (34) shows that in the presence of a frame adverbial, DA is excluded, while (35) shows that if a focused subject is moved to a higher position and has a definite reading, as is the case for the first person pronoun wiej, DA is again excluded.

32.	a.	Wiej binn-t /*binn-e den besten!
		We are the best!
	b.	Binn-e /*binn-t wiej den besten?
		Are we the best?
33.	dat/* darr-e [zölfs wiej] de wedstrijd wint
	that/that-Agr even we the game win
34.	dat/* darr-e [op den wärmsten dag van’t joar] wiej tegen oonze wil ewärkt hebt
	that/that-Agr on the warmest day of the year we against our will worked have
35.	WIEJ denkt Jan dat/*darr-e die pries ewönnen hebt, niet ZIEJ
	we think Jan that/that-1P.SG that prize won have, not they
	‘WE John thinks won that prize, not THEM.’
	(Hellendoorn Dutch, van Koppen 2012, p. 138)

Let us now have a look at the present accounts of CA in West Germanic. Because of the special properties of DA, two types of accounts have been proposed in the literature. For dialects of type A, where neither movement of a subject nor subject modification have an effect on the appearance of CA, it is assumed that CA is based on an agree relation between a C-head and a subject. For type B dialects, Fuß (2016) proposes a prosodic account in terms of string adjacency. In particular, Fuß (2016) assumes the post-syntactic movement of agreement features, which depends on strict string adjacency between the subject and Fin₀, accounting for the intervention effect of frame adverbials, as in (33) above. However, such an account cannot be extended to the dialects of type A (in which CA occurs in the presence of intervening elements), as was argued by Haegeman and van Koppen (2012), and the agreement-based account cannot be extended to dialects of type B. Thus, neither account can explain the phenomena in all of the diverse languages/dialects that display CA.

Before we sketch an alternative account that explains CA in both types of dialects, let me provide a description of the relevant data for CA in relation to the facts of subject-licensing in Cimbrian. It is immediately clear that the data in Cimbrian and CA in West Germanic exhibit very similar patterns that arguably call for a unified account. Let us thus consider whether the facts of CA can be explained in terms of anchoring a subject to a context. This implies that CA agreement should be seen as an alternative anchor (of the subject) such as clitics and da in Cimbrian.

First, I note that if the position of the complementizer (and the Wackernagel position) is identified with Fin⁰ in the C-domain, then an alternative anchor is expected since the subject is prevented, for syntactic reasons, from moving into [Spec, FinP]—since no element can precede a complementizer in Germanic—to anchor the clause in embedded clauses. For matrix clauses, as shown in (31) and (34), where a subject moves into or through [Spec, FinP], no alternative anchor is needed.

Furthermore, I note that CA in Hellendoorn Dutch appears to behave similarly to subject clitics in Cimbrian (rather than da) in being incompatible with focus. This is in line with the observation that an agreement morpheme in a type B dialect is related in form to subject pronouns. Finally, (34) illustrates that a frame adverbial in a type B dialect bleeds CA since frame adverbials, as I discussed above, allow for the anchoring of a subject without any clitic via binding.

In conclusion, I make the following proposal: by considering referential anchoring, a uniform syntactic account becomes feasible where the two types of CA are related by a diachronic process of reanalyzing movement as agreement (cf. Wratil 2016 on DA in Carinthian and Kansas Bukovina Bohemian) along the following lines:

(A) Type B dialects involve the movement of a subject clitic or of a correlate of the subject into Fin⁰ to anchor the predicate. This is why in cases of subject movement and subject modification, no overt CA morpheme appears. Dialects may then be taken to differ as to whether they allow subject clitics with focused subjects, as I propose is the case in Austrian Bavarian (see the data in (29) above), or whether they resort to a silent adnominal da. Dialects may not allow subject clitics with focused subjects, but they may also lack an adnominal da, and then the sentence is expected to be ungrammatical with or without CA, as is the case in Frisian (see the data in (31) above).

(B) Type A dialects are characterized by the loss of DA that can be explained in the present account in the following terms: the clitic is reanalyzed as verbal inflection, but the agree relation between Fin⁰ and the subject remains and is interpreted as feature evaluation, as specified in (36) with the consequence that no intervention or modification effect is expected and the agreement morpheme is assimilated (or identical) to the agreement morpheme on the finite verb.

36.	If term A agrees with function f (x), where x is a free variable, then x is evaluated
	with respect to the value assigned to A.

8. Conclusions

I have presented arguments showing that complementizer agreement is more than a quirky formal effect that appears in some West Germanic dialects, where we may ask what it is good for, and I have argued that CA can be taken to serve a purpose, namely, to anchor the subject in the context.

Furthermore, I have argued that the distribution of indefinite DPs and of strong and weak definite DPs in clauses follows from the assumption that these DPs must enter into a licensing relation with Fin⁰ in the C-domain. These arguments are based on the assumption that the context values of discourse referents and the values pertaining to the utterance and reference situations become accessible in this position. This approach thus throws new light onto phenomena such as clitic movement to a high position in clauses that occur in many languages, as well as onto the scrambling of definite DPs into higher positions in the middle field (that may be identified with positions in the lower C-domain) in German. These operations then cease to be quirky formal properties of these languages, and they can be taken, as with CA, as we argued in the previous section, to serve to license definite DPs in the context.