The interpretation of [+distal] in demonstratives and complementizers

3.1 Exclamatives

We first look at main clauses with overt complementizers, which in many languages can get an exclamative reading:^[10]

In these examples, the exclamative is only distinguished from a regular declarative sentence by the addition of the complementizer and a different intonation pattern. The intonation pattern alone is not enough for the exclamative interpretation. For instance, a sentence like Bio industry is still allowed!, with the same intonation pattern as the exclamative, still differs from an actual exclamative like (3a) in that it can be used to attempt to convince the Addressee of its propositional content. By contrast, the sentence in (3a) does not make an attempt at informing or convincing the Addressee of its propositional content, but actually presupposes it to be a shared presupposition in the Common Ground, and hence in the Shared Discourse Space. The use of the complementizer is therefore crucial for the interpretation as an exclamative. Zanuttini and Portner (2003) already showed that exclamatives are factive.^[12] On this view, exclamatives make reference to a proposition that is already presupposed in the Shared Discourse Space and relate a certain Speaker stance (surprise, anger, etc.) to it.^[13] We propose that the [+distal] complementizer in these exclamatives anaphorically refers to the presupposed proposition.^[14]

In these cases, the referent (the presupposition in the Shared Discourse Space) is always “far” from the Speaker: we do not find exclamatives with a complementizer or other grammatical marker that is specified for [−distal].^[15] The notion of Addressee involvement makes it easy to see why: if, following Zanuttini and Portner (2003), exclamatives require presupposition, they must refer to the Common Ground, and hence to the Shared Discourse Space. An exclamative cannot at the same moment introduce new, Speaker-personal information content into the discourse. As a result, the information content must be close to the Addressee, and therefore a distal element must be used.^[16]

3.2 The that/zero alternation in English object clauses

We can also use Addressee involvement to explain the alternation between overt and zero complementizers in English object clauses. Consider (7):

A common view is that the complementizer that in (7b) is “optional”, i.e., that its use is determined by style or register and that it does not have an interpretive value. However, the literature discusses many factors that can play a role in the choice between that and a zero complementizer. Two in particular suggest that we are actually dealing with an interpretively meaningful alternation and not with an entirely optional functional element.^[17]

Firstly, Bolinger (1972: 58) already noticed that the sentence with that in (7b) suggests some context between Speaker and Addressee. This context may be extralinguistic, as in the scenario he sketches:^[18] “Suppose you observe a stranger struggling to mount a tire. Feeling charitable you go over to him and say [7a]. Under these circumstances, [7b] would be inappropriate. But if the other person looks at you as if wondering why you came over, you might explain by saying [7b]” (Bolinger 1972: 58, example numbers adapted).

In the words of Bolinger (1972: 56), the complementizer still “reflects the demonstrative character of that” in that it refers to this shared context. After all, this use of that appears to be quite similar to the discourse deictic function of demonstratives (e.g. That’s a lie; Diessel 1999: 101). Both refer to some utterance, even though the utterance is only implied in (7b) (i.e., we assume there to be an implicit utterance along the lines of Why did you come over?). The situation is then quite similar to that of exclamatives: the use of an overt complementizer signals content in the Shared Discourse Space. Again, then, the referent (the presupposed utterance) is analyzed as “far” from the Speaker, triggering a distal element, because it is in the Shared Discourse Space, close to the Addressee. In (7a), no anaphoric element is present because the idea that the Addressee might need help has not yet been introduced, and is therefore not in the Shared Discourse Space. The presence or absence of the complementizer thus marks the presence or absence of shared context in the Shared Discourse Space.^[19]

The other relevant factor conditioning the choice between that and zero is that of subjectivity (Storms 1966: 262–265). Storms argues that sentences incorporating a that-clause are “less personal, less familiar, less warm, less friendly, less emotive” than their counterparts with zero complementizers (Storms 1966: 262). He gives examples from a witness interrogation in court, where sentences without that are used “to put the witness at her ease and at the same time to set an unsuspected trap” (Storms 1966: 263). Later, when it is important that objective facts are established, questions with that are used (Storms 1966: 264). Similar ideas appear in Wierzbicka (1988: 132–140), who relates that-clauses (as opposed to other complementation types) to knowledge. We believe that this subjectivity derives from the placement of the complement in or outside of the Shared Discourse Space. The lawyer cited by Storms (1966) uses that for propositions that are not yet in the Common Ground, but by using that he implicitly proposes to update the Common Ground to include them.^[20]

Previously, Kaltenböck (2006) already proposed to use the abstract notion of distance to explain the difference between that and zero in extraposed that-clauses (as in It is obvious (that) she did it), suggesting that the analysis could be generalized to object clauses as well (Kaltenböck 2006: 389 n. 20). In his view, the abstract notion of distance is interpreted as one or more of (a) illocutionary distance (asserting the complement with zero vs. disposing the matrix for illocutionary force with that), (b) temporal/anaphoric distance (using that for complement clauses whose content has already been talked about vs. zero for new information), and (c) emotional distance (à la Storms 1966). However, Kaltenböck (2006) does not provide a principled reason why old information should be distal (in terms of temporal/anaphoric distance); we might as well argue that discourse-old information is proximal, because what is close to us is better known than what is far. The notion of Addressee involvement provides an explanation: discourse-old information is distal because it is in the Shared Discourse Space, known and shared by the Addressee. Addressee involvement is also needed to explain the lack of a proximal complementizer this, which Kaltenböck’s (2006) analysis does not seem to predict. We return to this issue in Section 3.4. We discuss more related work in Section 6.2.

To finish our discussion of optional that in object clauses we briefly discuss the fact that that is required when the object clause is topicalized:

As discussed by Rizzi in Kratzer et al. (2020), there have been different syntactic accounts of this phenomenon. In this interview, Rizzi proposes an account that adopts an idea from Pesetsky (1994). Rizzi assumes that the lack of a complementizer indicates incorporation or cliticization of that complementizer into the selecting verb. Since the C head of the complement clause has already moved, the complementizerless clause cannot in turn move to a higher position: the complement clause is frozen in place. This would explain the pattern in (8). However, this account needs to introduce the otherwise uncorroborated assumption that complementizers incorporate into the selecting verb in English. By contrast, the analysis based on Shared Discourse Space that we present here suggests an explanation that derives from a wider generalization: that is required in topicalized object clauses because topicalizations are necessarily discourse-old, and hence in the Shared Discourse Space. The complementizer that is then required to indicate the shared status of the topicalized clause.

3.3 Overt root complementizers in Romance

Finally, presupposition and the Shared Discourse Space also play a role in constructions with a root complementizer found in several Romance languages. It may not be immediately obvious that the complementizers discussed here contain a non-proximal deictic element, but we return to this issue below. One type of root complementizer construction that we are interested in here has a sentence-initial adverb followed by an overt complementizer:^[22]

According to Etxepare (1997: 99 via Hernanz 2007: 166), the felicity of Spanish que in sentences like (9) is conditioned by the occurrence of a clear “linguistic antecedent” in the preceding discourse (cf. Etxepare 2010: 613). Thus, (9) is only felicitous after another Speaker has uttered a sentence like ¿Se ha enfadado Julia? ‘Did Julia get angry?’. We might analyze this linguistic antecedent as establishing a Question Under Discussion (QUD) in the Shared Discourse Space.^[23] The existence of this QUD then licenses the use of que. Thus, as in the cases discussed above, the use of an overt complementizer is licensed by the existence of an element in the Shared Discourse Space (namely, the QUD). Because we want to focus here on the properties of the Shared Discourse Space, we do not go into details about a possible formal representation of the QUD.^[24] What is relevant to us is only that the Shared Discourse Space (a) contains informational elements, (b) that these elements can be tracked and referred to by conversation participants, and (c) that some of these elements can be marked as being under discussion.

According to Kocher (2022: 75–82), the linguistic antecedent requirement is not as strict as assumed by Etxepare (1997, 2010, as que can also be used for future or hypothetical utterances:

In Kocher’s (2022) analysis, que merges in a high position in the left periphery in cases like (10), where it simply indicates that the sentence is subordinate. The Addressee can then infer that a verbum dicendi is implicitly understood. This is in contrast to cases like (9), where que merges in a low position in the left periphery where it expresses that a commitment to the proposition is attributed to the Addressee (an attributive feature in the sense of Poschmann 2008). This attributive feature, which is a form of Addressee involvement, explains why B’s response is felicitous in (11a) below but not (11b). In (11b), the attributive feature of que clashes with A’s sentence in which the proposition is described as a false belief:

We are agnostic towards the exact syntactic derivation leading to the attested surface structures. What is important to us here is that in both cases distinguished by Kocher (2022), the use of que is conditioned by the existence of shared context between Speaker and Addressee. In cases like (9), the proposition itself is placed in the Shared Discourse Space; in cases like (10), there is the salient verbum dicendi that provides the shared context.^[25]

We find similar constructions in other Romance languages:^[26]

In (12), Romanian că may only be used when the Addressee could have inferred the propositional content of the clause. Cruschina and Remberger (2017: 89) set up the following contexts. Suppose Ioana asks Alexandru if Ion will attend a conference next week. Ioana does not and cannot have this information, but Alexandru has spoken to Ion and knows that he is coming. Alexandru can then answer with Sigur va veni. However, suppose now that Alexandru does not have this information, but that both Ioana and Alexandru know that Ion is a big fan of the conference and would never miss it. In this context, Alexandru can answer with Sigur că va veni. The answer is then marked as an inference from information in the Common Ground between Ioana and Alexandru, rather than as private information of Alexandru. Again, we see that reference to the Shared Discourse Space (here in particular the Common Ground), and hence Addressee involvement, is marked by an overt complementizer.

A related phenomenon can be observed in Neapolitan. The following contrast is discussed by Sornicola (1996: 334–336) and Ledgeway (2011: 286–289):

On Ledgeway’s double subject analysis, chello/chillo is not a complementizer but a demonstrative, but it is still similar to the cases discussed above. In (13a), chillo is coreferential with the second subject (‘It has gone off, the heating’). In (13b), neuter chello cannot be coreferential with masculine ’o riscaldamento. The demonstrative must therefore refer to something else. It has “a distinctly explicative or adversative value, only proving felicitous in contexts that contain an implicit or explicit presupposition” (Ledgeway 2011: 287). We suggest that in (13b) the demonstrative refers to this associated presupposition, as is the case with the complementizers in Spanish, Catalan, and Romanian.

We have largely left French aside in the discussion above, because French que is nearly obligatory in all environments. There appear to be some varieties that do allow que to be dropped in some contexts.^[28] However, these cases have not been described in sufficient detail yet to be included in our discussion here.

We should pause here for a moment to reflect on the origin of these complementizers. Above, we argued that the [+distal] complementizer that marks the use of Shared Discourse Space because the Shared Discourse Space includes the Addressee and is therefore “far” from the Speaker. However, the Romance complementizers here are not demonstrative synchronically, so how does a [±distal] feature fit in? Note that these complementizers derive from Latin quod, which is composed of an interrogative element qu- and the originally neuter medial demonstrative id. Given the latter component, these complementizers do diachronically derive from a non-proximal demonstrative element. Furthermore, there is reason to believe that the interrogative element qu- is incompatible with proximity. For instance, consider that English has what from that and where from there, but not *whis from this or *where with an /i/-vowel from here. Rooryck (2003: 11–12) suggests that this is because something that is proximate to the Speaker is necessarily known to them. In this way the interrogative element qu- could also be seen as a [+distal] component, thus involving the Addressee.

Our analysis of these complementizers is very similar to that of exclamatives. In (12), the sentence without că has an “objective” interpretation (“It is certain that s/he’s coming”), whereas că triggers a “subjective, speaker-oriented” interpretation (“Of course s/he’s coming”), where the propositional content is inferred (Cruschina and Remberger 2017: 88–89). This Speaker-oriented interpretation uses că to refer to a presupposition, just as exclamative complementizers refer to the proposition presupposed by their complement. This is entirely in line with Gutiérrez-Rexach (2001: 184–186), who calls these sentences in Romance “evidential exclamatives” and analyzes them as in (14a):

According to Gutíerrez-Rexach, the evidential adverb claro ‘clearly, of course’ requires that its complement makes reference to some Question Under Discussion (QUD). For example, (14b) may be uttered if the Speaker has borrowed something from the Addressee and the Addressee has expressed doubts about getting it back. The QUD is topicalized by the complementizer que. Because the complementizer is demonstrative, the QUD does not need to be spelled out, but the complementizer does need to be overt. The complementizer effectively points to the QUD in the Shared Discourse Space. Note, however, that the fact that it points to the QUD (and not any other element of the Shared Discourse Space) appears to be a language-particular constraint: it applies in Spanish, Catalan, and Neapolitan, but not in Romanian, where că does not refer to a QUD but to any evidential basis for the claim made in the complement clause. What is at issue for us here is the generalization that the complementizer points to an element of the Shared Discourse Space.

Example (14b) illustrates the division of labor between the sentence-initial adverb and the complementizer, as well as the parallel with the exclamatives discussed in Section 3.1. As with exclamatives, the function of the complementizer is to mark the existence of Shared Discourse Space between Speaker and Addressee. The sentence-initial adverb in the Romance constructions only specifies the evidential interpretation.^[29] The [+distal] element is therefore again used to signal Addressee involvement.

3.4 Presupposition effects: summary

To summarize the findings from this section: evidence from a variety of constructions (exclamatives, English “optional” that, and the root complementizer constructions in Romance) suggests that the alternation between an overt complementizer with a [+distal] feature and a zero complementizer is related to presupposition. We explain this by suggesting that the complementizer refers to information content in the Shared Discourse Space. Distal elements are used in these complementizers because the Shared Discourse Space includes the Addressee, who is “far” from the Speaker. Note that the theory correctly predicts that we do not find [−distal] elements in these environments, that is, that there is no complementizer derived from the demonstrative this. These would correspond to presuppositions that are not shared with the Addressee; a contradiction in terms, since presuppositions are necessarily assumed to be shared by all interlocutors. The only available alternation is with a zero complementizer, which marks the absence of a presupposition from both the Shared Discourse Space and the Personal Discourse Space.

Our analysis raises questions for the traditional account of the grammaticalization of that and cognate complementizers. The traditional view is that that became a complementizer as a result of reanalysis of a cataphoric demonstrative: I say that: he comes > I say that he comes (e.g. Diessel 1999: 123–125; Roberts and Roussou 2003: 113–120). As a cataphoric demonstrative, that introduces new information, which would be consistent with the use in I say that: he comes. But the shift to I say that he comes would be odd if the complementizer that, as in our analysis, refers to Shared Discourse Space (as opposed to introducing new information).^[30] However, recent studies have suggested that the complementizer that instead developed from a correlative construction: I say that, that he comes (e.g. Axel-Tober 2017, and see Bate In preparation for a survey of finite complementizers in Indo-European). In such a construction, the first pronoun introduces new information but the second can be seen as referring to the Shared Discourse Space (as established by the first pronoun). This grammaticalization path therefore does not suffer from the same problem. Our analysis provides further support for this development.

Finally, although we focus here on complementizers derived from demonstrative pronouns, the phenomenon that finite complementizers are related to presupposition seems to be more general than that. For example, the Bulgarian relativizer deto (lit. ‘where the’, i.e. ‘the place where’) is also used to express Speaker stance about presupposed propositions: Săžaljavam, deto ne možax da dojda ‘I regret that I couldn’t come’ (Krapova 2010: 1240). It may therefore be that a finite complementizer does not need to be derived from a demonstrative pronoun, but that any deictic origin would suffice. We will not explore this further here.

6.1 Generalizing over sentential and nominal reference

We have proposed a unified analysis of the recycling of the proximal/distal distinction between the demonstratives this and that in terms of actual distance (the “distance” between deictic expression and referent) and Addressee involvement (the “distance” between Speaker and referent). This theory is also able to explain the correlation of this and that with direct and indirect speech reports, respectively, as well as the alternation between that (or a parallel finite complementizer) and a zero complementizer in a variety of contexts. These abstract distances are interpreted in different ways depending on the type of referent, as shown in Table 1, reproduced from the introduction.

The four types of environments discussed above have been categorized according to two binary properties here. First, our deictic elements refer to either information content or entities. We studied information content in Sections 2 (direct and indirect speech) and 3 (presuppositions), and entities in Sections 4 and 5 (exophoric and anaphoric demonstratives, respectively). Second, the well-known distinction between exophoric and anaphoric demonstratives for reference to entities generalizes to information content, where it distinguishes utterances from their meaning. Both exophoric demonstratives and speech reports refer directly to concrete things in the world (entities and utterances), whereas anaphoric demonstratives and the complementizers referring to the Shared Discourse Space refer only indirectly (to entities via linguistic antecedents, and to information content through a mental model of the discourse state).

There are two gaps in Table 1. First, actual distance (the “distance” between referent and deictic expression) does not seem to be used in anaphoric reference. We can understand why this is the case in the following way. In both exophoric and anaphoric reference there is a direct link between the referent and the Speaker, namely in the cognitive model of the Speaker. This allows the proximal/distal distinction to be recycled to mark Addressee involvement. But a direct link between the referent and the deictic expression, which is needed to describe actual distance, only exists in exophoric reference: in anaphoric reference, the link is indirect, through an intermediate linguistic entity. The fact that this link is indirect seems to make it difficult to interpret the distance expressed by the proximal/distal element in terms of the relation between referent and deictic expression in these cases, and therefore there is no actual distance there.

Second, proximal elements appear to be incompatible with anaphoric reference to information content (presuppositions): the complementizer that alternates with a zero complementizer rather than with a complementizer based on proximal this. This gap has already been explained in Section 3.4: using a [−distal] element in this type of reference would suggest that the Speaker refers to informational content that is new to the Addressee (because it is not in the Shared Discourse Space) without introducing it (because anaphoric reference is used). Such use of language would be incompatible with cooperative conversation. In other words, the analysis presented here explains why finite complementizers are so rarely derived from proximal demonstratives.

What unifies the interpretation of Addressee involvement in all four contexts is the fact that the referent is presented as accessible to, or tracked by, the Addressee. Depending on the context, this may have some further implications. This is particularly visible with demonstrative that, as shown in Section 4. In these contexts, that is in opposition not only with this (which would explicitly mark the referent as in the Personal Discourse Space of the Speaker) but also with the definite article the. The latter has no [±distal] feature, but can still be used in contexts where the referent is mentally accessible to the Addressee. Consider the following contrast:

Example (20a) can be used in a context where the Addressee is either already tracking the hammer in their Personal Discourse Space, or can easily identify it – that is, the already implies Addressee involvement. As a result, the meaning of that becomes more marked: excluding pointing contexts where actual distance is promoted, (20b) is most natural in situations where the Addressee is already tracking the hammer, not in situations where the hammer is only identifiable. We thus see that the interpretation of the Addressee involvement marked by [+distal] that becomes more marked when it enters into an opposition with the. Such an opposition is not available for the complementizer that. As a result, the interpretation of Addressee involvement is simpler in this environment, and is confined to referring to an element of the Shared Discourse Space.

6.2 Related work

In this paper we have sought to bring together a number of well-known and much studied phenomena in a single theory. We do not have space here to review the full history of scholarship of all these phenomena individually. However, work on some of these issues has, without relating them to the other phenomena, reached similar conclusions to ours, and therefore deserves discussion here.

In particular, there is a long history of work on so-called optional that in English, some of which has been referred to in Section 3.2. Of these, Yaguchi (2001) presents an analysis that is quite close to ours: her paper “elucidates the function of the non-deictic that by considering how the residual meaning of the demonstrative that is still in effect […] and what underpins the presence or absence of the non-deictic that from a cognitive perspective” (Yaguchi 2001: 1126). While Yaguchi reaches similar descriptive generalizations based on similar data, we believe the analysis needs refinement. In particular, Yaguchi (2001: 1127) describes the complementizer that as “non-deictic”, while she claims at the same time that it also preserves the function of the demonstrative to “deictically point”. It is unclear how the two can be reconciled. Furthermore, Yaguchi (2001: 1127) takes a leap by assuming that “non-deictic” that has to do with truth: “the use of demonstratives implicitly encodes the speaker’s presupposition that the hearer can identify the entity to which the speaker refers […] By the same token, non-deictic that […] signals that the speaker presupposes the contents of the complement clause to be referential, in other words, to contain true or valid information, whose validity can be proven by evidence.” We agree that that is referential in both uses and that this can entail presupposition, but stress that it is perfectly possible to refer to things that are not true or valid. This shows, for instance, in (7) above, where that is used to acknowledge an implicit question of the Addressee:

For this reason we have analyzed that using references to a Shared Discourse Space which, unlike the Common Ground, does not only contain presupposed propositions but also other information content, such as questions or rejected propositions, and entities. Yaguchi (2001: 1137–1139) discusses verbs like think, believe, and guess, which do not presuppose their complement, but does not use referentiality to explain the use of that with these verbs. Instead, the distance expressed by that would mark the greater amount of evidence and analytic thinking used to come to the conclusion stated in the complement. Yaguchi does not specify, however, how speakers choose between these different factors (referentiality and amount of evidence) when interpreting an instance of that. Furthermore, Yaguchi’s approach is problematic for verbs like doubt, which suggest that the Speaker favors presupposing the negation of the complement. Even if the use of that in I doubt that P has to do with the amount of evidence, it has to do with the amount of evidence for ¬P rather than for P. By contrast, in our account we analyze all these different cases as involving referentiality. For example, the use of that in I doubt that P reflects that the question whether P is the case is tracked by both Speaker and Addressee.

Dor’s (2005) position on optional that is also quite similar to ours: he suggests that “the predicates which can embed the bare clause, without the complementizer, are those which entail that a cognitive agent (in the majority of cases, their subject) has made an epistemic claim concerning the truth of the proposition denoted by the embedded clause” (Dor 2005: 347). This improves on Yaguchi (2001) since it accounts for verbs like doubt, though we would still widen the scope a bit to involve reference to questions (for which no truth claim has been made) as well. Furthermore, Dor (2005) is primarily descriptive and does not seek to explain why the use of that is related to truth claims. In our view (as in that of Bolinger 1972; Yaguchi 2001) this can be explained as a type of reference, and thus connected to the demonstrative that.

This brings us to related work on the similarities between demonstratives and finite complementizers highlighted in Table 1 above. Roberts and Roussou (2003: 111–116) dismantle a number of arguments for the supposed synchronic homophony of demonstrative and complementizer that, which is the basis for much of what we are doing here. Kayne (2014) argues that the complementizer that is still a demonstrative, but one that does not require “pointing”. He also addresses the question why this is not a complementizer, providing an explanation based on a first person feature as opposed to our [±distal]. This is compatible with our analysis if first person is seen as an interpretation of [−distal]. Most recently, Ritter and Wiltschko (2019) and Colasanti and Wiltschko (2019) have argued for a nominal Speech Act structure dominating the DP layer. As is well known, Speech Act structure on the CP level is used to mark the relationships between the propositional content and the Speaker and Addressee, thus formalizing the differences between declaratives, exclamatives, interrogatives, and other sentence types. On the DP level, the Speech Act structure would be used to express the relationships of the interlocutors and the described entity – in particular whether it is discourse-old or discourse-new. This formalization is readily applicable to the observations we have discussed in the present article.

6.3 Final remarks

By way of conclusion we want to discuss three final points. First, we wish to point out that paying attention to the fact that the two abstract distances are recycled in different ways depending on the type of reference allows us to resolve some apparent paradoxes. For instance, recall that Cheshire (1996) argued that the exophoric demonstrative that can express empathy with the Addressee:

On the other hand, Storms (1966) suggested that in the context of a witness interrogation, sentences without that are used “to put the witness at her ease and at the same time to set an unsuspected trap” (Storms 1966: 263). Thus, the demonstrative that in (15a) would engage with the Addressee, whereas it is the absence of the complementizer that does this for Storms (1966). By fleshing out what Addressee involvement really means in these different types of environments, the paradox can be resolved: Cheshire (1996) is talking about reference to entities, where the distal demonstrative establishes joint attention and hence empathy; Storms (1966) is talking about information content where Addressee involvement concerns the Common Ground, and hence the establishment of facts. In this way, Addressee involvement is a useful generalization from which other categories, such as empathy (Cheshire 1996) or “relating to knowledge” (Wierzbicka 1988) can be derived.

Second, a unified analysis of demonstratives and complementizers allows us to explain why that introduces finite complements rather than non-finite ones. Tsoulas (1996: 298) points out that the finite/non-finite distinction in clausal complementation can be better described in terms of “definite” and “indefinite” propositions. A proposition is definite when it uses a “definite” tense, that is, a tense that specifies a precise temporal point. In this sense, finite complements are “definite” and infinitival complements are “indefinite”; the latter can by their nature not be situated precisely in space. The selection of a tensed complement by the complementizer that can be derived from its demonstrative nature: it references the precise temporal point. In other words, the fact that the complementizer that takes finite complements is fully analogous to the fact that demonstratives are necessarily definite (in the common sense): both require their referent to be situated in space and time.

Finally, we might wonder where the relativizer that fits in Table 1 above. Its position is clearly in the lower left quadrant for anaphoric reference to entities. However, note that there is no [−distal] relativizer (the book *this/that is on the table here), which matches with the complementizer that in the lower right quadrant (anaphoric reference to information content). We can explain the lack of a proximal relativizer in the same way as we explained the lack of proximal reference to information content: since the referent/antecedent is mentioned in the immediately surrounding context, it is necessarily in the Shared Discourse Space and can therefore not be referred to by a proximal element. Therefore, although the relativizer that stands in the lower left quadrant, Addressee involvement is interpreted not as interaction/empathy with the Addressee (as with other anaphoric reference to entities) but using Shared Discourse Space (as with reference to information content). We thus find the distinction between overt and zero relativizers to be similar to that between overt and zero complementizers. For example, (21a) is uttered out of the blue by a detective sergeant to a responding officer, and the Speaker does not expect there to have been anything unusual. The relative clause thus does not have any grounding in space-time or previous discourse, and that can be omitted. On the other hand, suppose a customer is looking through the racks in a clothing store. The salesclerk may then ask (21b), where a zero complementizer would be odd: the fact that the customer is looking for something is presupposed. There is a well-defined set of items from which the answer can be drawn (all the clothes in the racks), in contrast to the open-ended nature of (21a).

In this paper we have analyzed a number of high-frequency uses of the proximal/distal distinction, but our discussion has not been comprehensive. It is expected that actual distance and Addressee involvement can be interpreted differently in other contexts. What we do commit to is the position that the proximal/distal distinction is interpreted in terms of the distance between Speaker and referent (Addressee involvement) and/or deictic expression and referent (actual distance). In this way, the present paper provides an instrumentarium for further analysis of other kinds of reference.