2.1. Sentence type

According to the philosophical literature (e.g., Russell 1919), two basic types of copular sentences can be distinguished: predicational and identity sentences. The function of predicational sentences is to provide more information about some individual via predication. In such sentences the postcopular phrase NP2 predicates a property of the subject referent introduced by NP1; cf. (7). Equative sentences, such as (8), also called identity sentences, equate the referents of the two NPs flanking the copula. Higgins (1979) introduces two additional types. One of these types is of particular relevance in this paper – the type of so-called specificational clauses as in (9).

Specificational clauses differ from the basic types in two respects. First, intuitively, in specificational clauses the NP2 specifies the “value” of the description given in the precopular NP1. Thus NP1in (9a) the president of the club sets up a variable (“the x that is the current president of the club at the given time”) and the NP2 John provides the value for that variable.

Second, specificational clauses in German have one peculiarity with respect to verbal agreement: the copula agrees not with the sentence-initial NP1 but with the sentence-final NP2. This can be seen when NP1 and NP2 do not match in number as in (10): NP1 is singular, while NP2 is plural.¹

Since in (10) the copula agrees with NP2, this NP can be assumed to be the subject and specificational clauses to have an “inverse” structure.

Third, it has been shown in the literature that specificational clauses have a fixed information structure, whereas the information structure of other types of copular clauses is not fixed in the same way (Heggie 1988, Heycock 1992, among others). Hartmann (2019) experimentally supports this observation: the postcopular NP has to be focused and the rest of the clause is the background. Thus, the sentence The culprit was John can be understood as being intended to answer the question in (11a). In the answer (11b), john corresponds to the wh-phrase of the question and is the focus, while the rest of the sentence is the background. NP1 cannot be focused even in an appropriate question context as in (12).

Thus, only the postcopular NP can bear focus in a specificational clause. It has been observed that the requirement of postverbal focus is typical for inversion structures in general (Birner 1996). For this reason, specificational sentences are considered inversion structures (Hartmann 2019, Heycock 2012, among others). According to this view, specificational and predicational clauses are derived from an underlying structure in which the copula takes a small clause consisting of a predicate-NP and a referential subject-NP, an idea that originally goes back to Heggie (1988). If the predicate-NP bears focus, the subject-NP moves out of the small clause into the sentence-initial position and the result is a predicational sentence with “canonical” order. A specificational sentence emerges if the subject-NP bears focus and the predicate-NP moves out of the small clause into the sentence-initial position, yielding “inverse” order.

The inverted agreement in German additionally supports the predicate inversion analysis in this language.

2.2. Discourse conditions on NP1

The assumption that specificational copula clauses are inversion structures has been further elaborated by Mikkelsen (2004). Following Birner (1996), she considers inversion an information packaging device that “serves a discourse connective function by letting material that is relatively familiar in the discourse appear before material that is relatively unfamiliar in the discourse” (Mikkelsen 2004: IX.) The condition for the placement of NP1 in the initial position of specificational clauses should then be that it is relatively familiar, i.e., more familiar than NP2. According to Mikkelsen (2004), the relative discourse-familiarity of NP1 qualifies it to be the topic of the specificational clause. But what kind of topic is it?

As will become clear in the next sections, NP1 is not an aboutness topic. According to Reinhart (1981), aboutness topics must be referential in order to create an address under which the information contributed by the rest of the clause could be stored. As I will argue below, NP1 is non-referential and denotes a property. Property-denoting NPs do not qualify as aboutness topics. Mikkelsen suggests, however, that properties can be assumed to be topics in the discourse-familiarity sense. To distinguish such non-referential topics in specificational clauses from aboutness topics, I will call them specificational topics.

What does it mean for a specificational topic to be discourse-familiar? Is pre-mentioning it enough? As a reviewer pointed out, the pre-mentioning of Genies ‘geniuses’ in the first sentence in (14) makes it given, but it doesn’t make its continuation, the specificational sentence in (14b), felicitous. The predicational sentence as its continuation is, however, acceptable.

Based on Prince (1992), Mikkelsen discusses different ways by which an NP in general and a specificational topic in particular can become discourse-familiar. The main idea of this discussion is this: the specificational topic is always related to some antecedent or “trigger” in the previous discourse or in the utterance situation. Although this relation may be different in each case, the familiarity/discourse-oldness of the trigger always projects to the whole specificational topic.

To examine the linking of specificational topics to preceding discourse, I conducted a corpus study. I collected examples from the newspaper corpus of the COSMAS database, the archive TAGGED C2, which contains morphosyntactically annotated newspaper texts from 2010‒2014 from German speaking countries. The search was restricted to newspapers from Germany. I searched for specificational copular clauses in which NP1 was accompanied by an article and analyzed the first 160 examples (supplementary file 1). My study supports the view of Mikkelsen that specificational topics are related to a discourse-old trigger. In the collected examples the trigger is introduced in the previous discourse or in the same clause. I identified three types of relations to the trigger: (i) bridging, (ii) inclusion, and (iii) identity. In what follows, I will look at these relations in turn. First, consider the specificational clause in (15).

In (15) Ziel is the head of the complex NP1 containing a genitive modifier dieser Wettbewerbsreihe ‘of this series of competitions’. Ziel is relational and unique: there is an associative possessive relation between the series of competitions and its unique goal – the series of competitions has one particular goal.² The NP1 das Ziel is clearly discourse-new since it is not used previously in the text. However, the demonstrative dieser in the modifier DP dieser Wettbewerbsreihe indicates that the Wettbewerbsreihe has been introduced in the previous discourse and is hence discourse-old. As Mikkelsen (2004: 213) observes, in such a case the NP can inherit discourse-familiarity from its trigger/possessor. The examples in (16) show that the trigger/possessor of specificational topics must be discourse-old. If it is discourse-new, i.e., indefinite, the construction is not felicitous:

Note that the trigger need not be explicitly mentioned as part of the NP1. It can be evoked from the utterance situation or be inferred from the context, as in the next example.

In (17) das Ziel is associated with ein ganzheitliches Konzept in the previous clause by an associative possessive relation, such that the goal is interpreted as being the goal of that clinical concept.

In the above examples (15/17), the trigger introduces a referent and NP1 requires an inference from the trigger on the basis of background knowledge. The relation between the trigger and the NP1 can be captured as a possessive relation in a very general sense. This type of discourse link between the trigger and a unique discourse referent has been dubbed bridging (Clark 1975).

Another type of relation I have identified in my corpus sample is the inclusive or sub-part relation. An example of such a relation is given in (18), where the (implicit) trigger of eine Person and die andere Person is a group of people implicitly introduced in the question, and the NP1 gets a partitive interpretation.

Since the trigger, the group of people, is discourse-old, the NPs eine Person and die andere Person inherit the discourse-familiarity from it. This example differs from the others discussed before not only in the inclusive relation to the trigger but also in the prosodic properties. The NP1 bears a contrastive stress and a rise contour (or complex fall-rise contour). The rising accent on NP1 die andere Person and the falling accent on the last constituent Erwin Rüddel in the comment build the so-called “Hutkontur” (literally ‘hat contour’). The rise contour on NP1 indicates a contrast to contextually invoked alternatives. These are characteristics typical for contrastive topics (Büring 2003, among others). Thus, the partitive interpretation of NP1 in specificational sentences is an indicator of a special type of specificational topic, namely the contrastive specificational topic.

The third way in which NP1 can be linked to its discourse-old trigger is the strongest one – the direct anaphoric relation: the potential referent of NP1 is identical to the referent of the trigger.

The NP1 der heutige Artikel der Financial Times is clearly discourse-old, as it is anaphorically related to the NP einen neuen Artikel in the previous sentence.

The literature on specificational clauses (Mikkelsen 2004: chapter 8, Heycock 2012: 219) points to the fact that specificational topics need not be definite but may be realized by NPs with indefinite articles too. Heycock claims that only strong indefinites, which generally have also been called “specific indefinites”, can occur in this position. According to her, the strong reading of an indefinite in English can be favored by different modifications, including the use of one rather than a and modification by a relative clause. Although in German the indefinite article and the numeral one have the same form, the specific interpretation may be triggered by the prosody: indefinite NPs can have a focus accent on the article and be realized as a contrastive topic as shown in (20a). The use of the stressed ein with the NP Ursache ‘reason’ here evokes the existence of other possible reasons, and a partitive interpretation of the indefinite NP with an inclusive relation to the trigger emerges. Another use of indefinites as specificational topics is illustrated in (20b). Here the article is unstressed.

I follow von Heusinger (2002) in the assumption that the referent of a specific indefinite, not only in its partitive reading as in (20a), must always be linked by a contextually salient relation to some already established trigger ‒ to the speaker or another discourse item. In (20b) the salient relation – the relation of worrying about – is introduced in the relative clause. It explicitly anchors ein Thema ‘a topic’ to wir ‘we’. Ein Thema is discourse-new, while uns is discourse-old. The whole NP inherits the discourse-familiarity of its trigger. This case is similar to the case of bridging described above. The difference is that here the trigger of the relation is introduced not in the previous discourse but as the modifier of the NP itself. What is also interesting in (20b) is that the anchoring relation is explicitly mentioned. All in all, we can conclude that specific indefinites used as specificational topics, like definites too, are anchored to a discourse-old trigger.

To sum up, specificational clauses are inverted structures with the focused subject in the clause-final position and a topic-NP in the sentence-initial position. This structure is remarkable since, according to standard assumptions, subjects tend to be topics and to occur sentence initially. Inversion of the predicate NP1 is subject to the discourse condition that it must be discourse-old. This condition can be captured as follows:

2.3. The referential type of NP1

In this subsection I want to discuss the referential status of NP1. In specificational clauses, the postcopular NP2 refers to an individual; the referential status of the sentence-initial NP1 is, however, controversial. Depending on the semantic type assigned to the NP1, different types of analyses of the specificational clauses have emerged. I will discuss two types of analyses. Some linguists assume that NP1 is individual-denoting and since NP2 is also individual-denoting, the whole sentence is an equative one. Other linguists assume that NP1 is a predicate and the whole sentence is an inverse predicational sentence. I will evaluate these analyses.

3.1. Empirical observations

To identify the types of nouns that are used bare as specificational topics, I conducted a corpus study in which I collected specificational copular clauses with a bare NP1 from the newspaper corpus of the COSMAS database. For the search the archive TAGGED C2 containing morphosyntactically annotated texts was used. The search was restricted to four years 2010-2014 of the supra-regional newspaper Die Zeit (cf. supplementary file 2). The search query was a copula sentence of the type [bare NP1 … ist … NP2 …].⁵ The automatic search yielded 5.741 copula sentences. After analyzing the matches, I identified 158 sentences as examples of specificational clauses. The examples were collected (cf. the supplementary file) and analyzed. The examples contained 85 different bare nouns as specificational topics. While 57 of them were used only once, the other 28 were used more frequently. The list in (31) presents these nouns with the frequency given in brackets.

In the next step I analyzed how NP1 fulfills the condition of discourse-familiarity in terms of a link to a discourse-old trigger. In Section 2.2 we identified three ways an implicit link to a discourse-old trigger can be established: by an anaphoric identity relation to an antecedent, by an associative binding relation to a possessor, or by an inclusion relation to a group antecedent. In my sample I found neither examples of the anaphoric relation nor examples of the inclusive relation. All examples of bare specificational topics for which a broader context could be found in the corpus display the same pattern: the bare noun is used as a relational noun that relates its referent to some discourse-old trigger by bridging. The bridging relation between the potential referent of NP1 and the trigger can be characterized in a broader sense as a possessive relation. The possessor argument is sometimes introduced in the previous discourse as in (32) (possessor = ein Boom ‘a boom’) or as a modifier of the bare NP itself. The modifier can take different forms, e.g., as an adverb (32b) or an NP (32c/d), or it can be realized as the first part of a compound (32e).

To see that the relationality of the bare noun is one crucial condition for the bare use, we can compare the use of the inherently relational noun Autor with the non-relational Schriftsteller. As (33a/b) shows, both nouns can be used as specificational predicates with the definite article. However, only Autor can also be used bare, while the definite article with Schriftsteller cannot be omitted.

While the possessor argument of the relational noun Autor can be introduced in the genitive NP (der Autor des Buches), Schriftsteller seems not to have such an argument in (33b) since this noun is not relational (*der Schriftsteller des Buches); cf. Schwarz (2009: 129).⁶ Thus, relationality is one crucial factor in licensing bare specificational topics. However, my further analysis of the data suggests that relationality is not sufficient.

According to my intuition, the bare use of NP1 always triggers a unique interpretation. In my corpus sample I did not find any context which would trigger or support a non-unique interpretation. From this I conclude that uniqueness is the second precondition for a bare use of an NP as a specificational topic. With respect to uniqueness, the nouns in my corpus sample can be divided into two groups. Nouns of the first group such as Hauptstadt ‘capital’, Besitzer ‘owner’, and Vorsitzender ‘chair’ are clearly unique: in a default context, they assign to a given possessor only one entity (e.g., a country has, or “possesses,” normally only one capital city). Other nouns such as Aufgabe ‘task’, Schwerpunkt ‘important point’, and Sponsor ‘sponsor’ are not restricted in this respect and can apply to one or more entities related to a given possessor (e.g., one institution can have more than one task, a program can have more than one important point, etc.). Interestingly, if nouns of the latter group occur bare in the predicate position of a predicational clause, they seem to allow a unique as well as a non-unique interpretation.

The second clause in (34) triggers a partitive non-unique interpretation, so that in the first clause digitization is interpreted as being one of many important points of the program; an indefinite article could be added. What is intriguing is that if Schwerpunkt occurs as topic in a specificational clause as in (35), only a unique interpretation is possible. The second clause triggering a non-unique interpretation leads to incoherency.

The descriptive generalization we can draw at this point is that if relational nouns occur bare their interpretation with respect to uniqueness is restricted: as specificational topics they may only receive a unique interpretation. The shift from non-unique to unique may be implicit. Sometimes it is, however, triggered by combination with a uniqueness-establishing adjective such as zentral ‘central’, superlatives or by compounding with a nominal stem such as Haupt-, Kern-, or Grund- that forces a unique interpretation (Hauptkritikpunkt ‘main criticism’, Kernbestandteil ‘core component’, Grundvoraussetzung ‘basic condition’). Thus, nouns that occur bare as specificational topics need not be inherently unique. Uniqueness may be established in the context or emerge in the course of semantic composition.

I conclude from my empirical study that relationality and uniqueness are the two key features that license the use of bare NPs as specificational topics, cf. (36): the uniqueness licenses the bare use and the bridging to a discourse-old possessor based on the relationality of the noun qualifies it for use as a specificational topic.

If we compare this condition with the general conditions on specificational topics in (21), we will see that bare NPs as specificational topics are more restricted than NPs with articles. While specificational topics with articles can be discourse-old in several different ways, bare NPs can acquire discourse-familiarity only by bridging.

3.2. Bare NPs in predicational and specificational clauses

As we have seen in the previous sections, specificational clauses have been analyzed as predicate inversion structures triggered by the topicalization of the predicate-NP. In this subsection I will compare the restrictions on the bare use of topicalized predicate-NPs identified in (36) with those on non-topicalized predicate-NPs.

In German, predicate nouns can also occur bare in the postcopular position of predicational copular sentences such as (37a), where they are not topics but are part of the comment. Traditional literature (e.g., Duden 2005) explains the bare use of predicate nouns in such sentences by their well-establishedness and stereotypicality: sortal nouns denoting well-established social groups of people such as professions or nationalities occur bare; cf. the recent analysis of such bare nouns in Geist (2019) and the literature cited there. However, there are other occurrences of bare predicates which cannot be explained with well-establishedness. Nouns such as Thema ‘subject’, Hauptziel ‘main goal’, and Grundproblem ‘basic problem’, which are unique and relational, may also occur bare; cf. (37b). Furthermore, Zobel (to appear) observes that nouns such as Bestandteil ‘component’ or Mitglied ‘member’, which are relational but non-unique, may also occur bare in the postcopular position (37c).

As can be seen in (38), not all NPs that can occur bare in predicational clauses can also occur bare in specificational clauses.

As I argued above, bare predicate-NPs can serve as specificational topics only if they obey three conditions: they must be unique, relational, and the trigger or bearer of the relation must be discourse-old. Obviously bare predicate-NPs such as Arzt in (38a) and Bestandteil in (38c) don’t fulfill these conditions: The predicate-NP Arzt is neither unique nor relational. The predicate noun Bestandteil is relational but not unique. The predicate noun Thema, however, fulfills the conditions on bare specificational topics: it is unique, relational, and its trigger can be discourse-old. However, even Arzt and Bestandteil can occur as a bare NP1 if they are modified by an adjective or word formation contributing to unique interpretation, and a relation to an old discourse trigger is explicitly or implicitly established in the context; cf. (39a/b).

One question remains: Why can’t relational nouns such as Bestandteil or Mitglied occur bare as specificational topics? They can occur in this position only if combined with a definite or indefinite article. The definite article would trigger a unique interpretation, and the indefinite article a non-unique partitive one; cf. (40a)

But why isn’t the bare NP in (40b) able to express partitivity? I do not have a straightforward answer to this question now. However, one possible explanation could go in the following direction: In Section 2.2 we have seen that only indefinites which are specific can occur in the initial position of specificational clauses and receive a partitive interpretation. Specific indefinites are strong and for that reason cannot be used bare. The article drop and bare use would “weaken” the form of such NPs and, possibly, prevent them from being specific.

To conclude, different types of predicate-NPs can occur bare in the postcopular position of predicational sentences. Only one type of them, namely NPs with the features [+unique, +relational], can also occur bare as a topic in specificational sentences because only such NPs obey the condition on bare specificational topics identified in (36).

3.3. The concept type of bare specificational predicates

According to the condition on bare specificational topics in (36), only NPs which can bear the features [+ relational, + unique] are felicitous predicates in specificational clauses. To identify the class of NPs with such features and to distinguish them from other types, I use the classification of nominal concept types proposed by Löbner (2011). He introduces a classification of nouns according to the concept they inherently denote. The concept types are identified by two binary features: [U] for inherent uniqueness and [R] for inherent relationality. The four basic noun types correspond to the four logical types, cf. Table 1.

The meaning of sortal nouns is sortal concepts that characterize their potential referents in terms of properties. The meaning of individual nouns is individual concepts that assign a unique referent to every appropriate context of utterance. Relational nouns characterize their referents in terms of a particular relation to some other object, usually called a possessor. Relational nouns proper are non-unique, the relation between the possessor and the referent is not one-to-one; for example, for the relational noun sister (of) in a given context of utterance and for a given possessor there may be any number of potential referents. Functional nouns combine the features of uniqueness and relationality. The meaning of functional nouns is functional concepts that characterize their referents in terms of a particular one-to-one relation to the possessor. For the functional noun father (of), for example, in a given context of utterance and for a given possessor, there may be only one potential referent who is the father of that possessor.

Löbner identifies a correlation between the concept type of the noun and the type of the determiner. According to his theory of determination, only inherently unique concepts are available for definite use and only non-unique concepts are compatible with indefinite reference. However, determiners of each type can be combined with nouns of each concept: if the restrictions of the concept type are not met, the noun is shifted to an appropriate concept type. On this basis, congruent and incongruent determinations can be distinguished. For example, Vater ‘father’ primarily denotes a functional concept, it is relational and unique. The definite determiner der Vater preserves the functional concept type. If Vater is combined with an indefinite article yielding ein Vater, the uniqueness (and in this case also relationality) is canceled and the nominal concept is changed into a sortal one. Löbner (2011) uses statistical, typological, and historical evidence to corroborate the hypothesis that incongruent determination is a marked option and involves a type shift of the noun.

My analysis so far suggests that bare NPs licensed as specificational predicates are unique and relational, and hence are functional concepts. I also assume that the concept of the noun can be modified by adjectives and word formation, as was shown above, thus the concept type of the NP emerges in the course of semantic composition.

What about nouns denoting concepts other than functional ones, e.g., individual nouns? Individual nouns such as proper names, pronouns, or nouns like Mond ‘moon’ and Wetter ‘weather’ do not occur as specificational subjects in my corpus sample. Interestingly, the latter nouns are not appropriate specificational predicates even if they are combined with an article. The exclusion of individual nouns in this position is expected since according to Löbner (2011) such nouns are inherently individual-denoting (type e) or individual concepts (type ⟨s,e⟩). Such nouns are excluded in a position where a predicate is required.

What about relational non-unique nouns such as Bestandteil ‘component’? As we have seen above, if a noun preserves its relational concept and non-uniqueness, it cannot serve as a bare specificational topic (cf. ex. 38c). In an appropriate context, however, some relational nouns such as Schwerpunkt can be shifted to a functional concept (ex. 35 above).

As to sortal nouns, they are neither unique nor relational and hence do not obey the condition on bare specificational topics. However, conceptual shifts seem to be possible even with such nouns. One interesting example of such a shift is (41).

Assuming that Bühne ‘stage’ is primarily a sortal noun, in the given example it is used as a functional concept “die Bühne des Ensembles” (‘the stage of the ensemble’) and means something like ‘the area of activity of the ensemple’. Interestingly, such a meaning shift and the drop of the definite article seem not to be licensed as easily if Bühne occurs in the postcopular position; cf. (42).

This suggests that topicality per se supports a shift of the non-unique sortal concept to a unique one. In a predicational clause where Bühne is not a topic, such a shift is harder to get. At this point it is not particularly clear why a concept shift from [‒unique] to [+unique] is easier for topics than for non-topics. I leave this issue for further research. What this example suggests is that, in principle, almost every noun can denote a functional concept in an appropriate context and be qualified for a bare use as a specificational topic. As a conclusion of the discussion in this section, we can formulate the following generalization:

Before we come to the analysis in the next section, I would like to summarize my findings so far. In this paper I analyze noun phrases that occur as a predicative complement of an inverted copular sentence, called a specificational sentence. To occur in the sentence-initial position the predicate-NP must qualify as a topic of a special kind. The topical status of the NP requires its potential referent to be related to a discourse-old trigger. Three types of relations to the trigger can be identified: identity, a bridging/possessive relation, and inclusion. The topic of specificational clauses may be realized not only by NPs with definite and indefinite determiners but also by bare NPs. Only NPs that can be interpreted as a functional concept, i.e., are [+relational, +unique], may occur bare in this position. Their relationality is a precondition for topicality: if the possessor argument of the NP is discourse-old, a bridging relation can be established. On the other hand, the uniqueness of NP1 seems to be a precondition for the omission of the definite article.

4.1. Two levels of definiteness

Recent research on definiteness in different languages suggests a distinction between an anaphoric and a uniqueness-based notion of definiteness. This distinction can be made explicit by two different forms of the article (e.g., the weak and strong article in Fering), or no determiner in one of the two cases (e.g., in Mandarin) (Schwarz 2009, 2013). Anaphoric definites (realized in German by a so-called strong form of the article) refer to objects that have been introduced in the discourse. In contrast, unique definites (realized in German dialects⁸ by the weak form of the definite article) pick up a referent which is the unique or maximal element in the property denoted by the NP. Cheng, Heycock & Zamparelli (2017) map the anaphoric/unique distinction to the semantic distinction individuals vs. singleton/maximal properties. This distinction is in turn mapped onto two distinct syntactic levels, a high DP, with a pronominal index i (argumental), which I will call the a(naphoric)DP, and a lower DP, which I will call the u(niqueness-based) DP⁹ (predicative or argumental). The authors assume that the heads aD and uD may be filled or empty. Even if the head is empty, the projection can still be semantically active and visible for interpretation, if material of the appropriate type moves to its head or specifier.

The definite article enters the syntactic representation in uD and signals uniqueness. Since the uDP resists having an antecedent, the existence of an individual having the property denoted by the NP is not presupposed. In argument positions, uDPs can be turned into the argument type e via the free type-shifting operator MAX, which takes singleton/maximal properties and returns their unique/maximal element.

The aD-layer can be made visible for interpretation by the head-movement of the definite article from uD to aD; cf. Figure 1. aDPs always have an antecedent in the discourse.

Figure 1

Definiteness at the syntax-semantics interface (based on Cheng, Heycock & Zamparelli 2017).

The semantic analysis of definites by Coppock & Beaver (2015) goes in a similar direction. They claim that the basic use of definite expressions is their predicative use and the basic contribution of the definite determiner is to indicate uniqueness without indicating existence. The existence can be contributed by an operator which shifts the predicate noun into the unique satisfier of that predicate. To support the assumption that predicative definites are non-referential and lack an existence presupposition, Coppock & Beaver use cancelation tests. They apply such tests to predicate-NPs in predicational sentences like (44). Example (44) shows that the existence of the ambassador can be canceled.

The next test of Coppock & Beaver (2015) examines the status of uniqueness. It shows that unlike existence, uniqueness cannot be canceled.

To capture the observation that the definite article primarily contributes uniqueness and not existence, the authors suggest the representation in (46), where ∂(|P| ≤ 1) codes the uniqueness presupposition.

According to this representation, the definite article picks out the singleton set of (or the property of being the unique) P if there is just one and the empty set (or empty property) otherwise. The definite article preserves the predicative status of the NP and only adds a uniqueness requirement. Going back to Figure 1, the representation in (46) can be assumed to capture the semantic contribution of the uniqueness-based definite article as an exponent of the functional head uD.

The commonality in the analyses of unique definites by Coppock & Beaver and Cheng, Heycock & Zamparelli (2017) is in a nutshell the following: in both analyses uniqueness-based definites are treated as singleton properties. They are semantically and syntactically more basic, and a free type shift to an argument type is possible. The definite article of uniqueness-based definites has the semantic contribution captured in (46). Cheng, Heycock & Zamparelli additionally take into account the fact that in some languages the article as an exponent of the functional head uD may have a weak or zero form, as was discussed by Schwarz (2013), among others. The capturing of all these characteristics of uniqueness-based definites in one model is attractive and will be used in the analysis of bare specificational topics in the next subsection.

The analyses discussed above were developed for definite NPs occurring in the postcopular position of predicational sentences. The question we will answer in the next subsection is whether precopular predicate-NPs in specificational sentences behave similarly and can be analyzed in an analogous way.

4.2. Bare specificational topics as uniqueness-based definites

The task is now to consider how bare NPs used as specificational topics fit the representation of two levels of definiteness in Figure 1. First, we have seen that bare NPs in the position of specificational topics are predicates that require a unique interpretation. This profile is the hallmark of uDPs. Second, according to my corpus study presented in Section 3, bare NPs as specificational topics are not used as direct anaphors to refer to an object introduced in the previous discourse. Example (47) illustrates this fact: the bare NP Universitätsrektor cannot pick up the antecedent in the previous clause.

The analysis of the corpus examples in Section 3.1 has also shown that bare specificational topics do not pick up a referent from a group of referents introduced in the previous context either. Does this mean that bare NPs have no existence presupposition? To test this I applied the cancelation test used in the study of Coppock & Beaver (2015). Most informants accepted sentences like (48). Thus, the existence of the referent of the specificational predicate can be canceled.

Since specificational predicates with an optional definite article, although they are unique, have no existential presupposition, Coppock & Beaver’s (2015) analysis for unique definites can be applied to them. For the uniqueness-indicating determiner, I assume the representation in (49) proposed by Coppock & Beaver (2015).

This determiner realizes the head uD. It combines with an NP denoting a functional concept with the features [+unique, +relational] and forms an uDP that can serve as a specificational topic cf. Figure 2a. But how can we account for the fact that unique specificational topics can also be used without an article?

Figure 2

Uniqueness-based DPs as specificational topics: Two alternative realizations.

In this case, I assume that the head uD with the feature [+unique] is empty, but the uDP projection is still semantically active. The empty head is licensed by the phrasal movement of the NP with the feature [+unique] into [Spec,uD] – an option discussed by Cheng, Heycock & Zamparelli for the licensing of functional projections with zero heads Figure 2b, cf. Figure 2a.

The next feature of NP1 we have to account for is relationality: the noun has a possessor argument, to which its potential referent is linked by a possessive relation. To fulfill the condition on specificational topics, the possessor argument must be discourse-old. The representations in (50) capture the two features of specificational predicates, uniqueness and relationality, as follows.

(50a) represents the meaning of the NP with the head noun Grund denoting a functional concept, while (50b) represents the meaning of the NP with the head noun Bühne denoting a sortal concept that is shifted to a functional concept. For the representation of uniqueness of predicate nouns in (50) I use Coppock & Beaver’s (2015) representation for definite articles. In order to account for relationality I assume that the predicate poss establishes a possessive relation between the referent of the noun x and the possessor argument y. According to (50), Grund denotes a singleton set of the reasons for y and Bühne denotes a singleton set of the stages of y, if there is just one and the empty set otherwise. If Grund or Bühne are used as specificational topics, the possessor argument gets an anaphoric index that indicates that it must be pre-established in the context; cf. (51).

According to (51) the bare noun encodes the same uniqueness as the uniqueness-indicating definite article. This NP may be combined with the definite article in uD or be moved to [Spec, uD] and be used as a bare uDP.

That uniqueness-based definites crosslinguistically may occur bare or be combined with a reduced form of the article has already been shown in the literature (cf. Schwarz 2013 and the literature cited there). In standard German, the article reduction is attested in prepositional phrases: the reduced definite article may be contracted with the preposition; cf. (52a) and footnote 8. Schwarz speaks of “weak article definites” in this case. As (52b) shows, an absolute reduction of the article to zero is not possible in this case. If weak article definites are an instance of uDPs, the question arises why the weak article cannot also be reduced to zero, as in the case of predicate uDPs; cf. (53).

The reduction of the unique article to zero in standard German seems to be syntactically restricted. Note that the function of the definite article in German, besides marking definiteness, is to mark morphological case. Prepositions assign case to the noun phrase and the article makes this assignment visible. This is the reason why in PPs the definite article is generally not reduced to zero, but only to a weak form that can still indicate case. I think that the reason for zero realization of the weak definite article with predicate noun phrases is that such phrases occur in the predicate position, which is a syntactically ungoverned position. In predicational copula clauses, the copula, which has been assumed to be the bearer of merely grammatical features, takes a predicate complement but does not assign case to it (Lang 1999, Geist 2013). The predicate NP occurs in its default form, which is the nominative singular. Since no case is assigned to the predicate noun phrase, there is no need to indicate case and the article need not be realized.

All in all, these considerations seem to indicate that weak article definites in PPs and uniqueness-based specificational topics can be considered to be two instances of the same category of uniqueness-based DPs and a unified analysis of definite descriptions involving uniqueness is possible, although I will not attempt to apply my analysis to weak article definites here.

The last remark concerns the scope of the analysis. The proposed analysis has been developed to capture the syntax and semantics of specificational topics under unique/non-anaphoric interpretation. As we have seen in Section 2.2, (ex. 19) some specificational topic-NPs with definite articles can serve as direct anaphors in the discourse. This behavior is more typical for anaphoric definites. The limits of this paper do not allow me to provide a detailed analysis for such cases and I have to leave this topic for another paper.

To conclude, bare NPs used as specificational topics are an instance of uniqueness-based definites. The article with such definites indicates uniqueness. In German it can have a reduced form, which is still able to indicate morphological case assigned to the noun phrase. Since bare specificational NPs as topics are already unique and are not assigned case by the copula, the definite article is redundant and can be dropped. This is possible because the bare unique DP encodes the same uniqueness as the unique definite article, which can still be optionally realized.