3.1.1. Anaphoric demonstratives

Our classification of anaphoric demonstratives differs from existing taxonomies that consider anaphoric and discourse deictic demonstratives as conceptually distinct classes (e.g., Diessel, Reference Diessel1999; Himmelmann, Reference Himmelmann and Fox1996; Levinson, Reference Levinson and Horn2004). In these taxonomies, anaphoric demonstratives are proposed to track discourse entities with a past and a future. They refer to ‘pre-existing’ or ‘pre-activated’ discourse entities, established in previous discourse via a nominal antecedent, which then often persist in subsequent discourse. Discourse deictic demonstratives, on the other hand, are typically argued to ‘point’ at a variety of non-nominal stretches of preceding discourse, thereby connecting this information to the current sentence. As such, they are usually considered ‘impure’ deictic devices: they do not refer to clearly definable pre-existing entities, but create entities on the spot, which rarely persist in subsequent discourse (e.g., Diessel, Reference Diessel1999; Hauenschild, Reference Hauenschild, Weissenborn and Klein1982; Himmelmann, Reference Himmelmann and Fox1996; Lyons, Reference Lyons1977). Although we acknowledge this distinction, and consider the type of antecedent (nominal vs. non-nominal) indeed as a crucial analytical variable, we propose to attenuate the conceptual distinction between nominal/anaphoric and non-nominal/discourse deictic. This position is inspired by the behavior of demonstratives in our corpus and congruent with ideas proposed elsewhere (see e.g., Cornish, Reference Cornish2001; Ehlich, Reference Ehlich, Jarvella and Klein1982; Kolhatkar et al., Reference Kolhatkar, Roussel, Dipper and Zinsmeister2018; Piwek, Beun, & Cremers, Reference Piwek, Beun and Cremers2008).

First, we consider demonstratives with nominal and non-nominal antecedents both deictic, as their referents need the deictic force of a demonstrative (rather than a regular pronoun or definite NP) to be accessed properly. In case of nominal antecedents as in example (1), demonstratives more often have a non-subject antecedent than regular pronouns (e.g., Brown-Schmidt, Byron, & Tanenhaus, Reference Brown-Schmidt, Byron and Tanenhaus2005; Çokal, Sturt, & Ferreira, Reference Çokal, Sturt and Ferreira2014; 2018; Fossard, Garnham, & Cowles, Reference Fossard, Garnham and Cowles2012; Kaiser & Trueswell, Reference Kaiser and Trueswell2008), and bring less accessible entities into focus (e.g., Ariel, Reference Ariel1990; Gundel et al., Reference Gundel, Hedberg and Zacharski1993, Reference Gundel, Hedberg and Zacharski2004; Hauenschild, Reference Hauenschild, Weissenborn and Klein1982; Linde, Reference Linde and Givón1979). In their function as determiner in a (modified) demonstrative NP accessing highly activated referents, as in example (2), they create additional inferences compared to definite determiners (e.g., ‘predicating’ demonstrative NPs, Cornish, Reference Cornish2011; de Mulder, Reference De Mulder1998; Doran & Ward, Reference Doran, Ward, Gundel and Abbott2015; Gaillat, Reference Gaillat2016; Goethals, Reference Goethals2013; Kleiber, Reference Kleiber1991; Maes & Noordman, Reference Maes and Noordman1995; Schnedecker, Reference Schnedecker2006). In the case of non-nominal antecedents and higher order abstract referents, as in example (3), demonstratives are found to be used more frequently than pronouns (e.g., Gundel et al., Reference Gundel, Hedberg and Zacharski2004; Kolhatkar et al., Reference Kolhatkar, Roussel, Dipper and Zinsmeister2018; Maes, Reference Maes1997). Demonstratives also enable access to new referents (e.g., recognitional that or indefinite this as in examples (7) and (8)) more easily than definite NPs, and more easily allow cataphoric relations, as in example (4), than pronouns. Finally, they are also more powerful in creating mental representations of referents based on indirect cues, as we will see below: both nominal cues (e.g., in deferred/bridging or generic reference, Doran & Ward, Reference Doran and Ward2019; Lücking, Reference Lücking, Coniglio, Murphy, Schlachter and Veenstra2018) and non-nominal cues (e.g., discourse deixis, Cornish, Reference Cornish, Nilsen, Amfo and Borthen2007; referent coercion, Kolhatkar et al., Reference Kolhatkar, Roussel, Dipper and Zinsmeister2018; Webber, Reference Webber1988). All these observations justify them to be considered a combination of deixis and anaphora, captured adequately by the notion of anadeixis (e.g., Cornish, Reference Cornish2001, Reference Cornish, Nilsen, Amfo and Borthen2007, Reference Cornish2011; Ehlich, Reference Ehlich, Jarvella and Klein1982).

Second, we consider demonstratives with nominal and non-nominal antecedents both anaphoric, in the sense that they always require some explicit antecedent trigger being present in the text. This condition is widely accepted for anaphors with nominal antecedents, but the ‘required presence’ of a non-nominal trigger should also be seen as a necessary condition for the construction of abstract referents (e.g., Kolhatkar et al., Reference Kolhatkar, Roussel, Dipper and Zinsmeister2018; Recasens, Reference Recasens2008; Webber, Reference Webber1988).

Third, anaphoric demonstratives with nominal antecedents in our corpus do not typically enable access to discourse referents ‘with a past and a future’. Most of these antecedents (78.6%) actually refer to abstract entities, they are often derived from verbs as in example (1), and rarely refer to persistent referents. This makes them conceptually similar to the referents accessed by non-nominal antecedents. In many cases, it is even unclear whether the antecedent is nominal or non-nominal, as we observed in our second coding exercise. For instance, in example (13), we may consider the whole NP (the fact that…) or only the that-clause as the nominal or non-nominal antecedent respectively. Whatever (subjective) interpretation we take, in either case, the demonstrative is roughly referring to the same conceptual entity.

Finally, we include in the two anaphoric classes regular cases as well as borderline cases, as we see this as the best starting point for more in-depth analyses of these cases. Scholars from different backgrounds use different labels for different types of borderline cases. Nominal antecedents are said to enable bridging, deferred, indirect, and/or inferrable reference (e.g., Gundel et al., Reference Gundel, Hedberg and Zacharski2004; Prince, Reference Prince, Joshi, Webber and Sag1981a), while non-nominal antecedents have been claimed to allow coercion, ostension, or indirect reference (e.g., Hedberg et al., Reference Hedberg, Gundel and Zacharski2007; Kolhatkar et al., Reference Kolhatkar, Roussel, Dipper and Zinsmeister2018; Webber, Reference Webber1991). Here, we will review some borderline cases as we found them in the corpus.

Borderline cases of demonstratives with nominal antecedents are mostly characterized by anaphor referents having a different semantic interpretation or denotation compared to their antecedent referent. Typical cases involving quantificational shifts, often discussed in semantic accounts of demonstratives, were not found in our corpus (e.g., “ Every dog in the neighborhood, even the meanest, has an owner who thinks that that dog is a sweetie”; Roberts, Reference Roberts, van Deemter and Kibble2002, p. 93). Shifts between a generic and a specific referent interpretation (Bowdle & Ward, Reference Bowdle, Ward, Ahlers, Bilmes, Guenter, Kaiser and Namkung1995; Doran & Ward, Reference Doran and Ward2019) are illustrated in the shift from a specific (a story) to a generic (these stories) referent in example (14), or the other way around (elephants → this one [elephant]) in example (15).

We coded a small number of demonstratives (n = 16) as bridging inferences (e.g., Apothéloz & Reichler-Béguelin, Reference Apothéloz and Reichler-Béguelin1999; Doran & Ward, Reference Doran and Ward2019), most of them having distal forms (n = 12) and/or coming from the book reviews (n = 13). Example (16) presents a fairly standard case based on a general knowledge inference (the “Our Father” in Polish → those words). Most other cases, however, partly also rely on genre-specific knowledge, in particular, on assumed knowledge of the book reviewed, and thus also have a displaced exophoric flavor, as in example (17) where all those chapters refer to the substantial part of the story that is situated in a retirement home.

As noted in previous work, demonstratives can furthermore sometimes defer reference (“A car drove by. The horn was honking. Then another car drove by. That horn was honking even louder,” e.g., Doran & Ward, Reference Doran and Ward2019; Lücking, Reference Lücking, Coniglio, Murphy, Schlachter and Veenstra2018; Wolter, Reference Wolter2006). In the corpus, we did not observe such cases, but we did find a productive class of deferred elliptic those/that anaphors (n = 156), as in example (18), where the demonstrative picks out the head noun of the antecedent NP (Cubans) to create a new entity (wet foot Cubans).

Finally, some demonstratives rely on information present in the previous discourse, but without representing a clear reference, as in example (19), where the demonstrative NP (that point) refers to a point in the story that is implied in the writer’s strategy of foreshadowing. These cases are often distal, like in semi-fixed expressions (e.g., at that, that being said, for that matter), and their semantic borderline type is hard to determine.

In sum, borderline cases of demonstratives with nominal antecedents show a varied picture. Yet, for all of them, the initial interpretation trigger can be found in the text, which is why we classified them all as anaphoric demonstratives.

For demonstratives with non-nominal antecedents, distinguishing between regular and borderline cases is much more complex. Apart from discussions on the nominal vs. non-nominal status of the antecedent, as in example (13), analysts have to find the exact syntactic extension of the antecedent, the semantic type of the anaphor, and the semantic type of the antecedent. In particular, the semantic interpretation turns out to be notoriously difficult, as both the antecedent and the anaphor can have a different semantic interpretation, to be selected from a fluid list of object types with increasing degrees of abstractness, including events, states, situations/facts, or propositions/speech acts (Asher, Reference Asher1993; Gundel et al., Reference Gundel, Hedberg and Zacharski2004; Kolhatkar et al., Reference Kolhatkar, Roussel, Dipper and Zinsmeister2018). Because of these complexities, early annotation attempts restricted themselves to distinguishing between direct and indirect cases of non-nominal demonstrative reference (Gundel et al., Reference Gundel, Hedberg and Zacharski2004; Hedberg et al., Reference Hedberg, Gundel and Zacharski2007).

More recently, Dipper & Zinsmeister (Reference Dipper and Zinsmeister2012) developed an annotation system based on three linguistic tests, applied here to example (13). The namely test identifies the antecedent in a namely construction following the demonstrative (this → this, namely (the fact) that the Amazon has no clouds above its channel near its mouth). The NP-replacement test adds a noun to the demonstrative, taken from a fixed list, and identifies the semantic type of the demonstrative anaphor (this → this state of affairs/situation/fact). The colon test adds a writer’s statement in front of the antecedent, and so identifies the semantic type of the antecedent (I state the fact: “the Amazon has…”). The combination of these tests allows these authors to detect semantic shifts between anaphor and antecedent with a fair level of intersubjective reliability, by using expert annotators and lists of nouns and statements connected to different abstractness levels.

Discussions remain however, as the linguistic tests do not prevent that several nouns or statements are acceptable at the same time. In example (13), the antecedent is clearly presented as a fact, but it remains unclear whether the anaphor has to be interpreted as a state, a situation, or a fact. Example (3) can be given a regular interpretation with the antecedent and the anaphor being both state-of-affairs or situations (i.e., Congress being asked to change the situation or state-of-affairs described in the antecedent sentence) or a slightly shifted one with an anaphor representing an action-event inferred from the antecedent (i.e., Congress being asked to carry out the action of changing the law which is responsible for the situation described in the antecedent). Also, the tests are not always easily applicable, for example in the semi-fixed construction in example (20), where the demonstrative refers to the container of the previous sentence or perhaps to the illocutionary act of saying, rather than to states, facts, or propositions expressed in the sentence.

In sum, congruent with our view, non-nominal antecedents are considered a necessary condition for the linguistic tests to be applied properly. Although the tests do not guarantee unambiguous coding of all subtle semantic changes between non-nominal antecedents and demonstrative anaphors, they are very useful in distinguishing regular from borderline cases.

3.1.2. Cataphoric demonstratives

Endophoric demonstratives also allow for cataphoric relations with nominal or non-nominal postcedents (see example (4)). Most taxonomies restrict these to proximal demonstratives with non-nominal postcedents (Chen, Reference Chen1990; Diessel, Reference Diessel1999; Gundel et al., Reference Gundel, Hedberg, Zacharski, Vargha and Hajičová1988; Himmelmann, Reference Himmelmann and Fox1996), although occasional distal cases have been reported as well (Danon-Boileau, Reference Danon-Boileau, Grésillon and Lebrave1984; Fraser & Joly, Fraser & Joly, Reference Fraser and Joly1980; Quirk, Greenbaum, Leech, & Svartvik, Reference Quirk, Greenbaum, Leech and Svartvik1985), and cross-linguistic differences in whether the proximal or distal (or any other) form functions as the preferred or ‘unmarked’ cataphoric demonstrative are expected. In our corpus, we found a small number of cataphoric demonstratives (n = 27). Six of them include a distal demonstrative, but all these can be considered borderline cases between an anaphoric and cataphoric interpretation, as in example (21) where the distal demonstrative can also be said to vaguely refer to the preceding proposition. The absence of cataphoric demonstratives with nominal antecedents in the literature and in our written corpus does not radically exclude such cases, as in example (5), but it is clear that they more typically fit in an interactive (and spoken) context.

3.1.3. First mention demonstratives

Apart from anaphoric and cataphoric demonstratives, we distinguish a variety of demonstratives (n = 193) used to introduce a new entity. We consider them text-based because the NP information following the demonstrative includes necessary triggers for an acceptable interpretation. In that sense, it is reasonable to consider them cataphoric as well, as has been done elsewhere (Deichsel & von Heusinger, Reference Deichsel, von Heusinger, Hendrickx, Devi, Branco and Mitkov2011; Gary-Prieur, Reference Gary-Prieur1998; Labrador, Reference Labrador2011, see also Kolhatkar et al., Reference Kolhatkar, Roussel, Dipper and Zinsmeister2018 for a similar use of the notion of cataphoric). In the corpus, their characteristics deviated from the general picture presented in Table 3: they are modified, often take up a non-subject function (81.9%) and non-initial (75.6%) sentence position, and are predominantly used in reference to first order (human or concrete) entities (88.6%), thus very much similar to the ‘predicating’ anaphoric modified demonstrative NPs.

An example of a first mention demonstrative that is extensively discussed in existing literature is recognitional that (e.g., Coniglio, Murphy, Schlachter, & Veenstra, Reference Coniglio, Murphy, Schlachter, Veenstra, Coniglio, Murphy, Schlachter and Veenstra2018; Cornish, Reference Cornish2001; Diessel, Reference Diessel1999; Himmelmann, Reference Himmelmann and Fox1996). The crucial characteristic of this class of first mention demonstratives is the assumption on the part of the speaker or writer that the addressee will be able to identify the intended referent on the basis of the newly provided information (Cornish, Reference Cornish2001; Gundel et al., Reference Gundel, Hedberg and Zacharski1993). Two additional modality-specific conditions are that the demonstrative is meant to signal to the addressee that a given referential expression may be elaborated on if necessary (Auer, Reference Auer, Hindelang and Zillig1981; Himmelmann, Reference Himmelmann and Fox1996; Schlegloff, Reference Schlegloff and Fox1996), and that the NP information is assumed to be privately shared between speaker and hearer (Diessel, Reference Diessel1999; Doran & Ward, Reference Doran and Ward2019). These two latter conditions hence do not apply in written communication where direct feedback is not available or audiences may be unknown. Therefore, in our corpus, recognitional demonstratives tend to stay more on the safe side and suggest only a generic type of familiarity, often combined with an attitudinal or rhetorical stylistic effect, as illustrated in examples (7) and (22). Note that the distal demonstrative in example (7) remains acceptable, even for readers not acquainted with the scene of the Royal Tenebaums, and that the effect of the demonstrative remains the same when reference to the new entity is repeated, as in example (22), the latter being observed in Norwegian as well (Johannessen, Reference Johannessen2008).

We should mention here that some scholars also discuss examples of unmodified recognitional or familiar that (“I couldn’t sleep last night. That dog (next door) kept me awake,” Gundel et al., Reference Gundel, Hedberg and Zacharski1993, p. 277; “[Sticker on rear window of car] Mind that child! He may be deaf,” Cornish, Reference Cornish2001, p. 300). Although one may see familiarity indeed as part of the pragmatic interpretation of these demonstratives, their acceptability also depends on generic scenario knowledge (dogs being part of neighborhoods; children being vulnerable in traffic), rather than individual knowledge about one specific dog or child. Moreover, the two cases have an exophoric flavor: the location of the rear window sticker activates the relevant exophoric context, and addressees with private knowledge of that annoying dog next door may give that dog a displaced situational interpretation (Himmelmann, Reference Himmelmann and Fox1996, p. 220–221). So, in non-interactive written discourse, where such conditions do not apply, we consider the presence of NP information crucial for these first mention demonstratives.

For indefinite this, as in example (8), roughly the same story holds. Proximal demonstratives in English are able to introduce new ‘indefinite’ referents (MacLaran, Reference MacLaran1980; Prince, Reference Prince and Cole1981b), but only if they can assure the reader that enough information is provided in the modified NP, and again with the demonstrative as a signal marking the upcoming new information, together with the pragmatic ‘proximal’ inference that the speaker or writer will provide this information. A typical context for these proximal NPs is the presentational sentence (e.g., She is this lawyer! Doran & Ward, Reference Doran and Ward2019, p. 250), but a similar effect can also be found in predicating anaphoric demonstratives as in example (23). In these instances, highlighting new information is combined with a positive evaluation based on the degree modifying effect of demonstratives (e.g., she is so good as a lawyer).

In our corpus, recognitional that and indefinite this were found almost exclusively in the book reviews (n = 23), attesting their informal nature. The distal cases (n = 15) represent typical cases of recognitional or familiar thatN. Only one of the remaining proximal cases could be replaced by an indefinite determiner, that is, in example (8). The others evoked similar inferences of familiarity based on assumed knowledge of the activated book, thus combining familiarity with a flavor of displaced exophoricity, as in example (24). All had a modified NP format, except example (25), where the reader of the book review is assumed to know that the referents of the distal and proximal unmodified demonstratives that barn, those years, and this man, as well as the referent of these poor men and women are to be found in the book, while those summers with my Grand-dad, a referent playing a part in the reviewer’s private life, again is modified with sufficient private (although not privately shared) information.

The most productive type of first mention demonstratives we observed are so-called restrictive that/those (n = 167) cases, found in all three text genres (news 75, wiki 13, and reviews 81): postmodified distal demonstratives without a head noun, as illustrated in example (6). Again, here the acceptability of the demonstrative type depends on the information provided by the subsequent relative clauses. These cases have a typical syntax and do not allow for replacement by a proximal variant. Semantically, they are somewhat related to quantificational demonstratives (Doran & Ward, Reference Doran and Ward2019; MacLaran, Reference MacLaran1980), and akin to deferred anaphoric demonstratives as in example (18).

3.1.4. In sum

We have proposed three main classes of text-based endophoric demonstratives. Anaphoric demonstratives, both in pronominal use and when part of a demonstrative NP, can have nominal and non-nominal antecedents, and include regular as well as borderline cases. Cataphoric demonstratives have a postcedent that is most typically non-nominal. First mention demonstratives are used to introduce new referents on the spot, thereby relying on information in the demonstrative NP. Objective formal variables (e.g., the presence of antecedent information preceding or following the demonstrative, or included in the demonstrative NP) suffice to distinguish the three main classes, as well as some subcategories, such as demonstratives with non-nominal antecedents or restrictive that demonstratives.

3.2.1. Origo demonstratives

Each communicative situation is intrinsically connected to a here-and-now (Bühler, Reference Bühler1934), which provides the opportunity to use situation-based demonstratives in reference to the origo of the ongoing discourse, as in example (9). This includes examples such as this week, this era, this room, or this country, which have previously been termed ‘symbolically exophoric’ (Levinson, Reference Levinson and Horn2004). These demonstratives are commonly observed in spoken discourse, but also occur in written text. In our corpus, the frequency of origo-based demonstratives differed substantially as a function of text genre: the news texts were responsible for 90% (n = 246) of all cases, consistent with the important role of space and time in news items.

Endophoric demonstratives can also refer in terms of a projected, transposed, imagined, or displaced origo (Bühler, Reference Bühler1934; Diessel, Reference Diessel1999; Levinson, Reference Levinson and Horn2004; Lyons, Reference Lyons1977). In our corpus, we however found them in these cases to be dependent on antecedent information needed to evoke a projected origo. Example (26) is taken out of the narrative part of a book review. The demonstrative has a clear antecedent, and replacement by a distal variant is a matter of stylistic effect rather than a drastic change of semantics (as would be in the case of origo deixis). In our taxonomy, this makes them text-based and therefore not situation-based.

3.2.2. Self-reference demonstratives

Each communicative situation also allows for self-reference via demonstratives, as in example (10). Such use of demonstratives includes deictic expressions referring to ‘the container’ rather than the content of the ongoing discourse, such as this chapter, this conversation, this article, this manual, etc. (e.g., Gundel et al., Reference Gundel, Hedberg, Zacharski, Vargha and Hajičová1988; Hauenschild, Reference Hauenschild, Weissenborn and Klein1982; Himmelmann, Reference Himmelmann and Fox1996; Paraboni & van Deemter, Reference Paraboni, van Deemter, van Deemter and Kibble2002). In our corpus, self-references (n = 112) were all NPs, almost all referring to the news reports (n = 102), as in example (10), and in exceptional cases also to the Wikipedia article (n = 3) or book review (n = 7).

Traditional taxonomies sometimes discuss borderline cases of self-reference when they distinguish between references to propositions (i.e., discourse deixis or ‘impure’ deixis) and references to the ‘material side of language’, i.e., [pure] text deixis (Cornish, Reference Cornish1999; Diessel, Reference Diessel1999; Himmelmann, Reference Himmelmann and Fox1996; Lyons, Reference Lyons1977; Webber, Reference Webber1991). In example (27), that word does not refer to the meaning of unsophisticate but to the container of the meaning, similar to example (20). Although we consider this a valid distinction, adding to the other subtle semantic distinctions found in anaphoric demonstratives, the presence of clear antecedent information makes these latter cases, as in example (27), according to our taxonomy, text-based and not situation-based.

3.2.3. Displaced exophoric demonstratives

Some demonstratives in the corpus enabled direct access to entities typically associated with specific written genres, and thus assumed to be activated on the basis of genre-specific knowledge. This class is particularly productive in the book reviews, as almost all reviews refer to this book as if it was physically present, as in example (11). These demonstratives can introduce the referent in a flexible way, using a pronoun only, as in example (30), a bridging inference as in example (31), or a generic demonstrative as in example (32).

We expect these displaced exophoric demonstratives to show up in many other genres as well. Manuals or product reviews, for instance, typically almost by default activate particular products or services. This results in a productive class of displaced exophoric demonstratives, reminiscent of displaced situational demonstratives (Himmelmann, Reference Himmelmann and Fox1996, p. 220–222).

3.2.4. Exophoric demonstratives

Situation-based endophoric demonstratives can paradoxically also be exophoric, for instance, when they refer to non-linguistic objects physically present in multimodal written genres, as via the words this diagram or this photograph, see example (12). They cannot be considered text-based, as the crucial cue for the interpretation of these demonstratives is not a linguistic one. On the other hand, they do not point at a genuine exophoric object either, as the referent can be seen as integral to the interpretation of the text. Therefore, we consider these cases endophoric situation-based exophoric demonstratives. In the corpus, due to the nature of the selected texts, they were not observed.

3.2.5. In sum

In the proposed taxonomy, situation-based demonstratives are demonstratives finding their interpretation in the origo of the discourse, in the container of discourse itself, in prominent (displaced) entities in the writing situation, or in non-linguistic objects present in multimodal written texts. None of them can reasonably be replaced by the distal variant in English and no linguistic antecedent is available. As a final note, it is worth mentioning that the usage most central in the study of exophoric demonstratives (this here vs. that there) did not show up in our endophoric corpus.