Discourse-related expletives: Challenges and opportunities

2.1.1 Topic-related expletives

Expletives in various languages have been related to topicality, notably in German, Icelandic, certain Italo-Romance varieties, Finnish and Kashmiri, as we outline in this section.^[3] Nonetheless, even within this group, the relevant elements appear to differ in terms of their motivation and thus do not necessarily form a homogeneous class, as we discuss below.

The most widely discussed ‘topic expletives’ in the literature are German es and Icelandic það (e.g. Lenerz 1985, Faarlund 1990, 63–6, Abraham 1993, Falk 1993, 163–4, Richards and Biberauer 2005, Håkansson 2017). These elements are generally compared against ‘subject expletives’ as in English and Mainland Scandinavian, which occur clause-initially and in non-initial subject positions, thus behaving positionally like subjects, e.g. (6) and (7).

By contrast, German es and Icelandic það are restricted to the clause-initial prefinite position and do not occur non-initially, e.g. (8) and (9).^[4]

Despite the reference to discourse, many authors account for Germanic ‘topic expletives’ on exclusively structural terms, as fillers motivated to satisfy verb-second (V2) (e.g. Haiman 1971, Lenerz 1985, Faarlund 1990, Richards and Biberauer 2005). In such work, the relevant elements are labelled as ‘topic expletives’ on the assumption that the clause-initial position they occur in is associated with topicality. In work that addresses the discourse connection more explicitly, these expletives have been claimed to signal the absence of a topic (e.g. Zaenen 1983, Theiler and Bouma 2012, Booth et al. 2017), given that they occur in thetic clauses (cf. Sasse 1987), i.e. an ‘anti-topic’ marking effect (cf. Jacobs 2001). Nevertheless, the exact reason why the grammar needs to mark a topicless construction in such contexts, and the precise mechanism by which this is achieved, is often not addressed (but see e.g. Fuß and Hinterhölzl 2023, De Cia and Cerullo 2024 as discussed below).

A similar discourse function has been recently attributed to two expletive elements in Italo-Romance, namely a in the Fornese dialect and chiru in the Cilentano dialect (De Cia and Cerullo 2024). Both of these elements occur in thetic clauses, e.g. in existentials/presentationals and impersonals (see (10) and (11)).

Crucially, if a nominal is topicalised and given a topical interpretation, and the sentence loses its thetic interpretation, a/chiru is ruled out, cf. (12-b) with (12-a).

De Cia and Cerullo (2024) argue that Fornese a and Cilentano chiru signal ‘zero aboutness,’ i.e., the absence of an aboutness topic in the utterance. This is essentially the same function as commonly attributed to the German/Icelandic expletives discussed earlier (see also discussion in Schaefer 2020, 11, ‘thetic expletives’).^[5] De Cia and Cerullo (2024) relate this to the claim in Erteschik-Shir (1999) that the truth value of the propositional content of all clauses must be checked against a topic (cf. also Erteschik-Shir 2007, 15). According to De Cia and Cerullo (2024), the two elements achieve this function via different mechanisms; while Cilentano chiru occurs in the canonical position for aboutness topics, Fornese a occurs in the canonical subject position. Given their different structural positions, one could argue that only Cilentano chiru is a topic expletive in structural terms; Fornese a, being a subject position placeholder, could instead be characterised as a subject expletive with additional discourse functions (Section 2.2.1).

Another relatively prominent topic-related expletive is Finnish sitä, a partitive form of the 3sg pronoun se (e.g. Vilkuna 1989, 58, 147, Holmberg and Nikanne 2002, Holmberg 2005). Holmberg and Nikanne (2002) show convincingly that sitä occurs in the immediately preverbal position, a dedicated topic position, lower than the position associated with contrast. Sitä always immediately precedes the element that is inflected for tense/subject agreement and can be preceded by maximally one (contrastive) constituent, e.g. (13), where italics in the English translation represents contrastive interpretation.

The Finnish topic expletive shows some striking differences with the Germanic and Italo-Romance topic-related expletives discussed thus far. While the German/Icelandic/Fornese/Cilentano expletives occur in clauses that lack a topic (‘anti-topic marking’), this is not the case for Finnish sitä. In genuinely topicless clauses in Finnish, the topic position can be unoccupied, resulting in a verb-initial (V1) clause (Holmberg and Nikanne 2002), e.g. (14). In such contexts, the postverbal element is interpreted as ‘coming into existence’ (Brattico 2020, 385) and cannot occupy the preverbal topic position (Holmberg and Nikanne 2002, 83).

Rather than being motivated in topicless clauses, expletive sitä occurs in clauses with a postverbal constituent which is a candidate for the topic position (Holmberg and Nikanne 2002), e.g. (15). Since minulle is a candidate for the topic position, cf. (15-a), V1 is ruled out here, cf. (15-b). A postverbal minulle is only grammatical if the preverbal topic position is occupied by the expletive sitä, cf. (15-c).

As such, rather than being an ‘anti-topic’ marker, as the Germanic and Italo-Romance expletives, the Finnish expletive could be referred to as having a ‘topic-blocking’ effect, occurring in the dedicated topic position in order to license a candidate for the topic position later in the clause.

Another element that has been labelled as a ‘topic expletive’ is Kashmiri yi (e.g. Bhatt 1994, 122, Kidwai 2004, 264). Yi is inserted in finite matrix clauses when the preverbal topic position is unoccupied (Bhatt 1994, 123, fn. 11), e.g. (16-a); the Kashmiri preverbal domain is characterised as Topic-Focus-V_finite (Kidwai 2004, 264). If the topic position is occupied by a referential topic, the expletive is ruled out, e.g. (16-b).

Interestingly, an additional discourse-related property has also been attributed to yi in such contexts, specifically that it signals that the clause takes as its ‘ground’ the default time/context, as defined by the discourse (Kidwai 2004, 242). As we discuss in Section 2.1.3, this appears to be a relatively common additional function of topic-related expletives.

The topic-related expletives discussed in this section indicate that such elements have diverse effects and motivations, from ‘anti-topic-marking’, i.e. marking the absence of a topic in thetic contexts (German es, Icelandic það, Fornese a, Chilentano chiru) to ‘topic-blocking’, i.e. occurring in a dedicated topic position to license a topic candidate later in the clause (Finnish sitä). Additionally, there is some overlap between topic-related expletives and situation anchoring (Kashmiri yi), as we discuss further in Section 2.1.4.

2.1.2 Focus-related expletives

Some expletives have been claimed to be linked to focus.^[6] For instance, Somali has been claimed to exhibit a focus-related expletive element, waxa(a) (<wax ‘thing’) (e.g. Livnat 1984, 89, Svolacchia et al. 1995, 78, Lecarme 1999, 296, Frascarelli 2010).^[7] Waxa(a) occurs in the preverbal dedicated focus position in a marked construction, whereby a heavy focussed constituent occurs outside the canonical focus position in a postverbal position, e.g. (17). We represent focus in the English translation via all caps; note that Somali does not have any prosodic marking of focus (Saeed 2004).

One possible explanation for the non-canonical, postverbal position of the focus in the waxa(a) construction is prosodic weight, given that it generally occurs with long focussed constituents (Svolacchia et al. 1995; Frascarelli 2010), cf. (17). Many authors acknowledge that waxa(a) has a focussing effect on the postverbal constituent (e.g. ‘cataphoric focus’, Svolacchia et al. 1995, 75), cf. (17). Indeed, without the presence of the expletive, postverbal positions for the focused constituent are ruled out, even if the more standard focus marker baa/ayaa is present. As such, the Somali expletive is an example of a focus-related expletive which is both structurally related to focus, since it occurs in the canonical focus position, and pragmatically related to focus, since it contributes a focus interpretation on other material in the clause. At the same time, it appears to be at least somewhat prosodically conditioned, occurring typically with long foci.

Another expletive element that occurs in a dedicated focus position is Basque ba- (e.g. Ortiz de Urbina 1994, Elordieta and Haddican 2016). In Basque, foci occur immediately to the left of the inflected verb. In contexts where no focus occupies the preverbal position, ba- must occur in that position, e.g. (18-a); V1 clauses without ba- are ungrammatical, e.g. (18-b).

Ba- is limited to such contexts; it is not possible in the presence of negation or foci that separate the verb from the left edge of the clause (Elordieta and Haddican 2016, 232–3), e.g. (19).

In other words, Basque ba- is motivated to fill an otherwise empty focus position. Given that ba- occurs specifically with synthetic verbal forms (cf. Ortiz de Urbina 1994, 127), Szendröi (2004, 218–9) claims that ba- has a prosodic motivation. Synthetic verb forms are stress-avoiding and, since the initial position is generally the main stress position in Basque, clauses with a fronted synthetic form, cf. (18), require something before it in the initial position to take up main stress. Ba-, Szendröi argues, fulfils this precise requirement. The claim that ba- is prosodically motivated in this way is further indication that focus-related expletives can be connected with prosodic requirements (cf. Somali waxa(a) above). At the same time, the prosodic properties of the Basque element are also striking because they challenge the general understanding that expletives cannot usually be stressed (cf. Sternefeld 2002, 294, Mycock 2004, 382).

Expletives that are structurally related to focus, acting as placeholders for a focus position, have also been observed in Romance varieties. For instance, Old Italian sì has been argued to be an expletive that occupies the position to the left of the inflected verb, i.e. the focus position (Benincá 1984, 1996, Poletto 2005).^[8] When sì (< Latin adverb sic ‘so/thus’) occupies this position, it is semantically vacuous, e.g. (20).^[9]

Another Romance expletive element, Sardinian ite (‘what’), occupies the focus position in questions if this position would otherwise be empty (Mensching 2016, 159) e.g. (21). Ite cannot co-occur with predicate and focus fronting, and the subject must be postverbal (Mensching 2016, 156).

Being a focus position filler in questions, ite shows similarities with wh-expletives, which occupy the wh-position in a matrix clause, cf. (4-b) (Section 1). On the basis of crosslinguistic data, Mycock (2004) states that wh-expletives occupy positions consistent with being focussed. Thus, on this view, wh-expletives can be viewed as a subtype of focus-related expletive.

Another expletive element that has been shown to relate to focus in a complex way is Hungarian azt (<distal demonstrative ‘that’.acc, cf. az ‘that’.nom), which occurs in the preverbal position in sentences with a clausal argument (de Cuba and Ürögdi 2009), i.e. in the dedicated focus position (cf. Kiss 1987). In a pragmatically neutral sentence, the presence of azt is conditioned by factivity (see also Section 2.1.5); non-factive predicates (e.g. ‘say’) feature the expletive azt in the preverbal position, while factive predicates (e.g. ‘regret’) are not possible with azt, e.g. (22).^[10]

However, when the clause argument is in contrastive focus (i.e. contrasted with other potentially relevant propositions), azt always appears, irrespective of the status of the matrix predicate with respect to factivity, e.g. (23). We represent contrastive focus in the English translation via italics.

Hungarian azt is thus an interesting example of an expletive that is both structurally related to focus, occurring in the dedicated focus position, but whose presence is also conditioned by focus, specifically by the status of the clausal argument with respect to contrastive focus.

In sum, like topic-related expletives (Section 2.1.1), focus-related expletives appear to be a relatively diverse class in terms of their specific motivations. While all of the elements discussed in this section are structurally related to focus, occurring in dedicated focus positions in the given languages, they differ in terms of the precise environments in which they are motivated. For instance, some are conditioned by certain prosodic requirements (Somali waxa(a); Basque ba-), some by the presence of interrogative force (Sardinian ite, wh-expletives) or, in the case of Hungarian azt, by factivity and contrastive focus.

2.1.3 Expletives and information status

In some languages, an expletive co-occurs with a clausal argument and marks the information status of the clausal argument, i.e. whether it represents given or new information.^[11] In Russian, for instance, the expletives ono and èto signal that the clausal argument is (at least partially) discourse-given (Pekelis 2019), e.g. (24-a). This is confirmed by the ungrammaticality of such elements when the clausal argument is informationally new, e.g. (24-b).

Similar observations have been made of German es as it occurs in sentences with a clausal argument, e.g. (25), where es co-occurs with a clausal object, e.g. (25-a).^[12] In (25-a), both speaker and hearer know about the arrangement to go to the cinema tomorrow expressed in the clausal argument; as such, es in such contexts has been claimed to mark the clausal argument as shared knowledge between speaker and hearer (Reis 1977, Berman et al. 1998). Further support for this comes from data like (25-b), where the clausal argument uncontroversially expresses new information, and where es is ruled out in the matrix clause.

In contrast to the traditional view on German es, Fuß and Hinterhölzl (2023, 44–5) point out that, while it can co-occur with a clausal argument that denotes a given situation, it can also occur with a clausal argument that introduces a new situation. Furthermore, elsewhere in Germanic, it has been observed that expletives co-occur with clausal arguments that express discourse-new information, as for Old Icelandic (Booth 2018, chapter 6), e.g. (26).

This suggests that the relationship between such expletives and the information status of the clausal argument in such contexts in Germanic is more complex than standardly assumed.

In sum, the data from Russian and Germanic show that expletives which co-occur with clausal arguments typically relate to the information status of the clausal argument later in the sentence. However, this particular discourse property is not only limited to expletives that co-occur with clausal arguments; for instance, in Paduan, the expletive subject clitic a occurs in clauses that convey new information, as in (27), or in exclamative contexts. As such, the expletive subject clitic a indicates the newness of the clause it appears in (cf. the discussion of Fornese a and Cilentano chiru in thetic, i.e. all-new clauses in Section 2.1.1).

2.1.4 Expletives and situation anchoring

Certain expletives have been claimed to relate to the anchoring of the clause they occur in, connecting it to the previous discourse or the speech act (‘situation anchoring’). As already mentioned in Section 2.1.1, this particular function appears to coincide often with topic-related expletives. For instance, Kidwai (2004, 264) claims that the Kashmiri topic-related expletive discussed already in Section 2.1.1 signals that the sentence containing it must take as its ‘Ground’ the default time/context (as defined by that particular discourse), cf. (16-a) above. Similarly, the German clause-initial prefinite expletive es, discussed as being topic-related in Section 2.1.1, has been claimed to serve as a situation topic in utterances that characterise a situation rather than an individual, linking the utterance back to the utterance situation (Fuß and Hinterhölzl 2023, 42–3). Fuß and Hinterhölzl (2023) claim that clause-initial es in German is inserted as a situation topic if the speaker intends the statement to be about a situation. While (28-a) is a statement about the individual Hubert Haider (categorical), (28-b) is a statement about a situation (thetic).

Unlike categorical judgements that can be anchored via an individual topic, the argument is that thetic judgements which characterise a situation need to be anchored via a situation topic (cf. discussion in Hinterhölzl 2019, 214–215). Given that a range of Germanic languages make use of expletives in thetic contexts (cf. Gast and Haas 2011), the situation anchoring function of the German expletive in contexts like (28-b) could potentially be extended to Germanic expletives on a broader level.

Romance also shows evidence for a connection between topic-related expletives and expletives which serve a situation anchoring function in thetic contexts. In Neapolitan, chello (<‘that (one)’, nt) has been described as an (invariable) expletive element that can only appear in thetic contexts like (29) when it is linked to the previous discourse and is not out-of-the-blue.^[13] Parallel to the Kashmiri and Germanic elements just discussed, it has been claimed that chello topicalises the situation referred to, also contributing an ‘explicative’ or ‘adversative’ value (Ledgeway 2010).

In Cilentano, a close variety to Neapolitan, a cognate element, cheru, shows similar behaviour (Cennamo and Cerullo 2021, De Cia and Cerullo 2024), cf. (30).

As already discussed in Section 2.1.1, Cilentano has a related element to cheru (nt), namely chiru (m), which appears in thetic clauses and has been recently argued to be a placeholder for an unoccupied (aboutness) topic position (De Cia and Cerullo 2024). The fact that cheru and chiru, with their distinct discourse properties, are related further reinforces the connection between expletives that are structurally related to topicality, and the situation anchoring function; see footnote 13 for a similar situation with chillo/chello in Neapolitan.

Elsewhere in Romance, Colombian Spanish expletive eso has also been claimed to serve an anchoring, situation topic function (Camacho 2016, 13–4). Evidence for this is that eso requires a presupposed discourse situation in order to be felicitous, e.g. (31-a); it cannot occur in a wide focus context like (31-b). Eso can occur both in main and embedded clauses, in a high position, preceding wh-elements (Camacho 2013, 49).

There is also another type of expletive that is linked to contextualisation, but rather than referring to the preceding discourse, it narrows down the contexts in which the sentence is appropriate. This is the case for, e.g. Vietnamese nó and West Flemish tet (Greco et al. 2017). In Vietnamese, the contextualisation effects appear to be speaker-related; the use of nó is only felicitious when the speaker has sufficient background information to report on the specific event, e.g. (32). Nó is homophonous with a third-person subject pronoun and appears in the preverbal subject position; being in a complementary distribution with preverbal subjects, it is incompatible with transitive and ergative verbs. Much like Fornese a, nó could be considered a subject expletive with additional discourse functions (cf. Section 2.2.1).

According to Dao (2021), this epistemic specificity should be reinterpreted as a corollary of the speaker’s involvement in her own discourse. The Vietnamese expletive-like element thus not only narrows down the context, but also marks the speaker’s attitude towards the proposition, which brings us to the last major type of discourse property which expletives relate to crosslinguistically, discussed next.

2.1.5 Expletives and speaker-related meaning

Expletives have also been claimed to be related to various types of speaker-related meaning, e.g. indicating the level of commitment and/or attitude of the speaker towards the content of the clause (cf. ‘exceptional expletives’, Greco et al. 2017). As well as the Vietnamese example discussed in Section 2.1.4 (cf. (32)), expletives that relate to the speaker have been described for a range of Romance varieties. In some non-standard Ibero-Romance varieties, the function of neuter or masculine expletive pronouns can be broadly characterised as marking an ‘emphatic epistemic value’ (Corr 2017, 56), although their precise interpretation differs per variety. For example, in Portuguese dialects, ele has been analysed as conveying speaker commitment and reinforcing the illocutionary force (Carrilho 2008), e.g. (33). Indeed, the use of ele is not felicitous when combined with an expression that mitigates the directive force, e.g. (34). Ele can co-occur with preverbal subjects (cf. (33)). This is evidence that it does not occupy the subject position but is located in the left periphery, occurring in both main clauses and complements of verbs of saying and perception.

Comparable uses of expletive pronouns are attested in Galician (el/il, expressing emphasis, Uriagereka 1995), Dominican Spanish (ello, an intensifying/exclamatory expletive, e.g. Hinzelin 2009, Gupton and Lowman 2013), seventeenth–nineteenth century Spanish (Silva-Villar 1996, Camacho 2013) and Balearic Catalan (ell, an emphatic or exclamative marker, e.g. Todoli 2002, 1370–1). Sardinian has borrowed ello/ellu(s) from Catalan, which is used in (rhetorical) questions and exclamatives (Remberger and Hinzelin 2023). The pragmatic use of the third person masculine pronoun illu/iddu in Sicilian and Calabrian, e.g. (35), seems to parallel the Ibero-Romance expletives, occurring mostly in questions and exclamatives (Ledgeway 2013, 281–2). However, its syntactic distribution and discourse values require further investigation.

Various expletives that are structurally related to topicality (i.e. occur in positions associated with topics, cf. Section 2.1.1) have also been argued to be related to speaker meaning in some way. For instance, Kaiser (2019) argues that the Finnish expletive sitä, which occurs in the dedicated topic position in the language (cf. Section 2.1.1), is also related to speaker/addressee meaning and indicates that the covert subject is associated with the speaker, e.g. the pair of impersonals (‘zero person sentences’) in (36). The example with the expletive in (36-b) is interpreted as conveying the speaker’s own experience (i.e. a type of evidentiality), from which it is claimed that sitä signals that the sentence expresses speaker-oriented affective meaning.

Similarly, the Icelandic topic-related expletive það, discussed in Section 2.1.1, has been linked to a rather different type of speaker-related meaning, namely the level of assertion on the part of the speaker (Rögnvaldsson 1983, Sells 2005). The evidence for this is that, in certain embedded clauses where það is optional (at least for many speakers), the more strongly a clause is asserted, the less felicitous the expletive-less version is, cf. (37).

Likewise, low acceptability is reported for það in presupposed (i.e. non-asserted) clauses such as restrictive relatives (Rögnvaldsson 1983).

In English, meanwhile, the presence of the expletive object pronoun it has been related to factivity (Kiparsky and Kiparsky 1971).^[14] For most speakers, it can co-occur with the complement clauses of factive verbs, but not of non-factive verbs, e.g. (38).

The reverse pattern is found in Hungarian, where the expletive azt – at least in neutral sentences – occurs only with non-factive predicates (cf. (22) in Section 2.1.2).

2.1.6 Summary

This section has outlined five major discourse properties which expletives have been linked to crosslinguistically, based on the work to date: (i) topicality, (ii) focushood, (iii) information status, (iv) situation anchoring and (v) speaker-related meaning. As the survey of these five aspects has revealed, many discourse-related expletives exhibit a degree of multifunctionality, relating to more than one of these properties. In particular, there is some evidence that expletives that are topic-related also often relate to situation anchoring (German clause-initial prefinite es, Kashmiri yi, Neapolitan chillo/chello), or are conditioned by factors concerning speaker-related meaning (Finnish sitä, Icelandic það). We regard examples of multifunctionality as reflecting a broader, more general property common to expletive elements, as we discuss next in Section 2.2.

2.2.1 Expletives, discourse functions and subject-/object-related functions

We have already seen in Section 2.1 that discourse-related expletives can themselves be multifunctional in serving more than one discourse function at once. In fact, many of the classic subject/object expletives from the literature have been claimed to exhibit multifunctionality, not only satisfying the need for a subject/object position to be filled, but also marking certain discourse properties. One relatively common example crosslinguistically is where an expletive element co-occurs with a clausal argument later in the sentence, cf. Hungarian (22)–(23), Russian (24), German (25), Icelandic (26) and English (38). In several languages, such expletives – which are standardly viewed as structural placeholders for the canonical subject or object position in the matrix clause – have been claimed to contribute discourse-related effects. This is, for example, the case with Russian ono/èto and (clause-internal) German es, which, when co-occurring with a clausal argument later in the sentence, mark that clausal argument as being given (Section 2.1.3). Such elements thus appear to serve a dual function, not only acting as structural fillers for the canonical subject/object position, but also contributing interpretive effects relating to information status. A similar duality of function is shown by the English expletive it, which co-occurs with an object clause (cf. Section 2.1.5), and whose presence, as discussed earlier, has been related to assertion and factivity. This small sample indicates that expletives which co-occur with clausal arguments often combine subject/object-related functions and discourse-related functions.

One might be tempted to address the fact that expletives which co-occur with clausal arguments can have discourse-related effects by assuming that such elements do not qualify as true expletives but are in some way referential. Indeed, many have previously claimed this (e.g. Bolinger 1973, Hoekstra 1983, Bennis 1986, chapter 2, Langacker 2011). However, this is not the only example of expletives where subject/object-related functions and discourse-related functions appear to coincide. In fact, some of the classic subject expletives have been claimed to have discourse effects, such as English existential/presentational there, cf. (6) (e.g. Rando and Napoli 1978, 308–9, Davison 1984, 826–7, Ward and Birner 1995, Basilico 1997, 282). For instance, Rando and Napoli (1978, 308) argue that the function of expletive there is to provide a ‘dummy topic’ in a sentence which would otherwise have none, while the later comment portion of the utterance is ‘more strongly emphasised or focussed upon’. Similarly, Bentley et al. (2015, 58–9) claim that non-agreeing subject clitics in existentials in northern Italian dialects contribute to the marking of the postverbal noun phrase as a non-topic, e.g. (39).

Thus, the information-structural role that has been claimed for various subject-related expletives in existentials/presentationals is similar to the topic-related expletives discussed in Section 2.1.1, which have been claimed to serve an anti-topic-marking function (i.e. signalling a topicless clause) (German es, Icelandic það, Cilentano chiru). Such parallels indicate that subject expletives relate to topicality in a non-trivial way, which we regard as a reflex of the fact that subjecthood and topicality often coincide, with the subject serving as a default topic in many languages (cf. Lambrecht 1994, 131–7, Andrews 2007, Bresnan et al. 2016, 100).

In terms of the borderline between discourse-related expletives and subject-related expletives, this is clearer to draw for some of the elements discussed in Section 2.1 than others. For instance, some of the elements can co-occur with preverbal subjects (e.g. Portuguese ele, cf. (33)), making it easy to rule out a subject analysis for such expletives. At the same time, Fornese a, Cilentano chiru (De Cia and Cerullo 2024), Vietnamese nó (Greco et al. 2017, Dao 2021) and Dominican Spanish ello (Camacho 2013, 45) are incompatible with overt preverbal subjects, suggesting that subject-related considerations are an important conditioning factor on the expletive. An example of the incompatibility of Dominican Spanish ello with a preverbal subject is provided in (40).

In sum, many of the classic subject/object expletives relate to discourse in a non-trivial way, such that it is often very challenging to try to delimit discourse expletives from subject/object expletives in a clearcut way. As such, instead of asking to what extent traditional subject/object expletives can be shown to relate to discourse properties, we regard a more suitable question for further empirical investigations to be to what extent expletives exist which are purely syntactic, and do not involve discourse at all. We see this as an important avenue for future research.

2.2.2 Discourse expletives versus discourse particles

If one allows, as we do in this article, for expletives to have discourse-related effects/motivations, this raises the question as to what extent such expletives can be viewed as distinct from (non-expletive) discourse particles. This particular issue is neatly illustrated by the differing views which have been put forward with respect to some of the Romance elements discussed in Section 2.1 (e.g. Portuguese ele, cf. (33)–(34), Dominican Spanish ello, cf. Section 2.1.5 and Balearic Catalan ell, cf. Section 2.1.5). With respect to the classification of such elements, previous contributions generally fall into two broad camps: (i) those that take them to be expletives on the understanding that the notion of expletives is broad and encompasses also non-subject expletives with discourse-related effects/motivations (e.g. Ledgeway 2002, 2013, Corr 2017), and (ii) those that argue that such elements are discourse particles rather than discourse expletives and thus not a challenge to the traditional view of expletives as purely syntactic elements that are subject-related (e.g. Hinzelin and Kaiser 2007, Kaiser 2009, Remberger and Hinzelin 2009, Sheehan 2016, Cognola and Casalicchio 2018). This discussion has been particularly prominent in the wider context of the status of expletives in NSLs (Section 3.2).

More broadly, the claim that certain languages exhibit topic/focus-related expletives (cf. Sections 2.1.1 and 2.1.2) is not in itself uncontroversial. One potential problem which all topic- and focus-related expletives share is the general assumption that only semantically contentful, referential expressions can be topics/foci (e.g. Brattico 2020, 387). Nevertheless, if one adopts the broad working definition of expletives as in (5), namely that they are invariable, non-referential, semantically vacuous elements that act as placeholders for specific positions at phrase structure, and if one also recognises that there are many discourse-configurational languages that have designated topic and/or focus positions rather than subject and/or object positions (cf. Kiss 1995), then the label of ‘topic/focus expletive’ does not seem unreasonable for many of the elements discussed in Sections 2.1.1 and 2.1.2.

Allowing for a broad notion of expletives that extends beyond subject/object expletives necessitates a way to delimit discourse-related expletives (e.g. a topic expletive) from discourse particles/markers (e.g. a topic marker). The particular distinction one draws here will depend heavily on one’s theoretical approach but, generally speaking, we see the structural nature of expletives as placeholders for specific positions at phrase structure as key here, and we assume that the discourse import of expletives relies on them occurring in a particular position at phrase structure, which would otherwise be unoccupied. This structural aspect is at least a rather robust way to single out discourse expletives from regular discourse particles/markers. As we outline in Section 4.2, the parallel architecture of Lexical Functional Grammar (LFG, Kaplan and Bresnan 1982, Bresnan et al. 2016, Dalrymple et al. 2019) allows for a possible solution along these lines.

3.2.1 Consistent NSLs

In Section 2.1, we saw various examples of expletive(-like) elements in consistent NSLs (CNSLs), e.g. Old Italian (13), Sardinian (21), Colombian Spanish (31), Neapolitan (29), Portuguese (33), Calabrian (35) and Dominican Spanish (40). Specifically with regard to these Romance expletives, two lines of argumentation can be found in the literature. On the one hand, some authors have argued that the presence of expletive pronouns in Romance forces us to broaden the notion of expletive (Corr 2017, 49), while others have seen such elements as cause to rethink the classic hypothesis as in Table 1 that null referential subjects implies a lack of overt expletive subjects (e.g. Camacho 2013, Ledgeway 2013). For instance, Camacho (2013, 20) concludes on the basis of data from Galician and Dominican Spanish that the connection between null referential subjects and null expletives is ‘not direct’, and that the availability of null referential subjects is ‘neither a necessary nor sufficient condition for null expletives’.

On the other hand, several scholars have argued that pronouns such as Dominican Spanish ello, Galician el, and Portuguese ele are not expletive subjects, but are rather discourse markers (e.g. Hinzelin 2009; cf. also Section 2.2). On this view, such elements do not constitute counterevidence to Rizzi’s generalisations (cf. Table 1) because they neither qualify as expletives nor as subjects. In fact, these expletives can co-occur with a preverbal referential subject, e.g. (41). Moreover, their distribution seems limited to root contexts and complements of verbs of saying/perception (Corr 2017, 52–3).

At the same time, the analysis of these elements as discourse particles/markers faces certain problems. First, in some Ibero-Romance varieties, the relevant element occurs in prototypical expletive contexts, including with existential and impersonal predicates, but not with transitive verbs (Corr 2017, 52–3). Moreover, Dominican Spanish ello cannot co-occur with preverbal subjects (Camacho 2013, 45), e.g. (40) (see also discussion in Section 2.2.1). Such limitations on their distribution are unexpected for discourse particles/markers. Second, the fact that such elements are pronouns in form is also relevant, since discourse particles more often derive from adverbs or adjectives (e.g. Cruschina 2010), while expletives very often emerge from pronominal forms (cf. the many examples throughout this article).

3.2.2 Partial NSLs (PNSLs)

PNSLs impose person restrictions on the omission of a definite subject pronoun, especially in third person root contexts (Roberts 2019, 207). Of the languages discussed in this article, Finnish and Russian have been argued to be PNSLs.

In Finnish, null subjects are only permitted in first and second person; in the third person, subjects need to be overtly expressed; otherwise, the subject will be interpreted as generic (a property typical of PNSLs, cf. Holmberg 2005, 2010, 92). The Finnish expletive discussed in Section 2.1.1, formally identical to the partitive form of the 3sg pronoun sitä, seems to occur only in third person, which matches up with the conditions on null referential subjects in the language, whereby only first and second person referential subjects can be null, but third person referential subjects must be overt. However, as shown in Section 2.1.1, the occurrence of sitä is dependent on discourse conditions concerning topicality rather than just grammatical person considerations. Another property that is often found in PNSLs is that, while they require the preverbal position (SpecTP) to be filled, it is not only the subject that can appear in this position and fulfil this requirement (Roberts 2019, 212). In Finnish, this preverbal position is a dedicated topic position (cf. Vilkuna 1989, 1995), and it is exactly this position where sitä appears. However, since it is not obligatory in every topicless clause, this EPP-like requirement can thus not fully account for the distribution of sitä (see also Greco et al. 2017, 72–4 for discussion).

Russian is another PNSL with discourse-related expletives (Pekelis 2019, 190) (Section 2.1.3). Like in Finnish, pro-drop is permitted only with the first- and second-person subjects but excluded with third-person subjects (Barbosa 2019, 490). Moreover, a high degree of topicality seems to be a requirement for subject omission (McShane 2009). The discourse-expletives èno/ono, however, are not related to topicality but rather to givenness (cf. Section 2.1.3). Thus, their occurrence does not seem directly linked to the (im)possibility of omitting subjects in Russian.

3.2.3 Radical NSLs

The last type of NSL are radical NSLs, also called ‘discourse NSLs’ (e.g. Huang 1984). These are languages that typically have no agreement marking on the verb and that allow the omission not only of subjects, but also other arguments, dependent on discourse constraints, e.g. topicality. Unlike in CNSLs, the use of an overt subject pronoun in radical NSLs does not trigger a special interpretative effect (Roberts 2019, 217). Vietnamese is an example of a radical NSL with a discourse-related expletive, nó (cf. Section 2.1.5 and Phan and Lander 2015, 400–1, for Vietnamese as a radical NSL). Nó occurs in what has traditionally been considered the classic (subject) expletive context, namely in the subject position when there is no preverbal subject. However, unlike canonical subject expletives, its presence is not obligatory and depends on discourse conditions, i.e. encoding speaker point of view (Greco et al. 2017) or ego-evidentiality (Dao 2021), which are not immediately reconcilable with those that govern pro-drop, e.g. topicality.

3.2.4 Summary

Discourse-related expletives, typically in subjectless contexts, are found in all types of NSLs (consistent, partial and radical). Whether the occurrence of such expletives poses a problem for the NSP and the predictions in Table 1 depends on their precise analysis (i.e. whether they are considered subjects or not), an issue that remains unresolved. The working definition of expletives we adopt in this article (cf. (5)) clearly allows for the possibility of expletives that fill positions other than dedicated subject positions and, as such, in principle broadens the class of expletives beyond specifically subject expletives. Indeed, the data discussed in Sections 1 and 2.1 in our view provide robust crosslinguistic evidence for a class of expletives that extends beyond strictly subject expletives. Assuming this broader notion of expletives, the presence of expletives in NSLs is not necessarily a challenge to the predictions in Table 1, provided that (i) Table 1 is assumed to hold only with respect to subject expletives and (ii) the relevant ‘unexpected’ expletives can be shown to not qualify as subjects.

As such, we see the need for more empirical research targeting the question of whether such expletive elements can/cannot be shown to be subjects. As it stands, in those languages that impose (grammatical or discourse) restrictions on the occurrence of null referential subjects, discourse-related expletives do not systematically occur in cases where these conditions are not met, i.e. where an overt subject would be required. The fact that the expletives are not complementary to the contexts where null referential subjects are permitted poses problems for a subject analysis of such elements. The relation between null subjects and and expletives is thus far from straightforward. If these elements are indeed non-subject expletives, then another important question for future research is how to theoretically capture the distinction between discourse-related expletives and discourse particles/markers (Section 2.2.2). This, of course, depends on one’s theoretical approach, as we discuss further in Section 4.